Inflammatory plasma proteins predict short-term mortality in patients with an acute myocardial infarction

This study was based on data from the population-based Augsburg Myocardial Infarction Registry which was established in 1984 as a part of the MONICA-project (Monitoring Trends and Determinants in Cardiovascular disease) and since then operated as KORA (Kooperative Gesundheitsförderung in der Region Augsburg) Myocardial Infarction Registry [10]. The study area consists of the city of Augsburg, Germany, and the two adjacent counties comprising a total of approximately 680,000 inhabitants. All patients aged between 25 and 84 years being admitted to one out of eight hospitals in the study area were consecutively registered. More detailed information on case identification, diagnostic classification of events and quality control of the data can be found in previous publications [10, 11]. For the present study blood samples were taken from patients with ST-elevation myocardial infarction admitted to the university hospital of Augsburg between May, 2009 and July, 2013 (blood from patients was collected solely in this hospital). Out of all patients from which blood samples were taken, biomarkers of 398 consecutive patients were measured (mean age: 63.5 years (SD: 11.9), male: 73.1%). For 3 patients there were missing values for the biomarkers, thus 395 cases were included in the analysis. All study participants gave written informed consent. Methods of data collection were approved by the ethics committee of the Bavarian Medical Association (Bayerische Landesärztekammer) and the study was performed in accordance with the Declaration of Helsinki.

Trained study nurses interviewed the participants during hospital stay using a standardized questionnaire. To confirm the information provided by the patients and to collect additional information, the patients’ medical chart was reviewed. Demographic data, data on cardiovascular risk factors, medical history, comorbidities (including diabetes), medication before and during hospital stay, as well as at discharge were collected from each patient. Furthermore, laboratory parameters including glucose measurement, ECG and in-hospital course were determined.

Between 2009 and 2014 plasma samples were obtained within the scope of cardiac catheterization, which was in general performed immediately after hospital admission. Right at the beginning of the catheterization, EDTA blood samples were taken (arterial blood). This was immediately followed by the processing of the blood samples in the catheter laboratory (centrifugation, aliquoting and freezing at − 80 °C).

Protein measurements of the 92 biomarkers were performed on plasma samples using the Proseek® Multiplex Inflammation panel, developed by Olink Proteomics (Uppsala, Sweden) and based on the Proximity Extension Assay (PEA). In the supplementary material, we give a brief description of the methods used for protein level quantification. Further information on the process of measurement can be found in a previous publication [12] and directly at the website of Olink Proteomics [13]. The selection of the 92 biomarkers of the inflammation panel was predetermined by Olink Proteomics and can´t be chosen individually.

All other blood parameters were measured in clinical laboratory at the university hospital of Augsburg during hospital stay of the patients as part of the regular diagnosis and routine treatment.

The endpoint used in this study was 28-day all-cause mortality. Information on mortality was obtained from the patients’ medical charts as well ascertained by regularly updates on the vital status of all registered persons of the MI registry with data from the population registries. Death certificates were obtained from local health departments.

For the comparison of categorical variables, Chi-square tests were performed and the results were presented as absolute frequencies with percentages. For normally-distributed continuous variables, Student’s t-tests were used. For continuous variables that were not normally-distributed we used nonparametric tests. The results are presented as mean and standard deviation (SD) or median and inter-quartiles range (IQR).

First, we standardized the values for each biomarker (we centered and normalized the variable so that the transformed variable had an expectancy value of 0 and a statistical variance of 1 for every biomarker). The standardization provides comparability between the 92 biomarkers. To examine the associations between the biomarkers and 28-day mortality, we calculated 92 logistic regression models, one for every biomarker. The first models were adjusted for sex and age. To control the effect of multiple testing, we FDR-adjusted the obtained p-values. In a subsequent step, we calculated the same logistic regression models and adjusted for sex, age, renal function according to GFR, diabetes mellitus, hypertension, hyperlipidemia and PCI (p-values also FDR-adjusted).

Based on these models, we identified biomarkers that were strongly associated with 28-mortality (FDR-adjusted p-value < 0.01). We added the standardized values of the selected biomarkers to obtain a summed value, which then was used as a prediction score. To assess the predictive ability of the score, we compared it to the established GRACE score [14] serving as a reference. As we did not have all information in the exact defined manner, we tried to replicate the score as best as possible (for details see supplementary material). There were two main deviations from the original score: no information on Killip class was available, but we had information on left ventricular ejection fraction (LVEF), which we used instead. Furthermore, as only ST-elevation myocardial infarction cases were included in this study, every case was assigned to the ST deviation group. There were cases with missing values for the variables LVEF, cardiac arrest, elevated heart enzymes and heart rate at admission. As just ignoring these cases could potentially cause a selection bias, we conducted multiple imputation by chained equations. The number of iterations was 5 and the number of created imputed data sets was 5 as well. The imputation process was performed with MICE-package (R statistic software). The subsequent analyses were calculated with the pooled results from the imputed data sets.

Subsequently, we calculated a combined score (GRACE score and biomarker score). Therefore, we standardized both scores (GRACE and the new developed biomarker score) and added both together. For the calculation of the biomarker score, 3 out of 398 patients had missing values. The missing values required to calculate the GRACE score were imputed (36 patients had at least one missing value for a required variable). Consequently, 395 cases were included for the ROC curve analysis of the GRACE score and combined total score.

The predictive ability of the three scores—the new biomarker score, the original GRACE score and combined total score—was compared. We performed ROC analyses and calculated AUC for each score and compared the results using bootstrapping. Finally, we calculated continuous net reclassification improvement (cNRI) and integrated discrimination increment (IDI) to further compare the predictive abilities.

While the interleukins IL-6 and IL-8 are classified as pro-inflammatory cytokines [15, 16]; IL-10 is suspected to have predominantly anti-inflammatory effects [17]. A systematic review by Kristono et al. summarized a potential association between inflammatory cytokines and long-term adverse outcomes in acute coronary syndromes, including studies analyzing the cytokines IL-6, IL-8 and IL-10 [18]. They concluded that some studies reported significant associations between individual cytokines (including IL-6, IL8 and IL-10) and MACE (major adverse cardiovascular events); but they also found considerable heterogeneity in the methods and results of the studies included their review [18]. They furthermore suggested that a combination of multiple cytokines might have a better association with MACE than individual cytokines [18].

Many potential pathophysiological pathways, which might convey a potential association between these cytokines and the outcome of AMI patients are described in the scientific literature. IL-6 for example is more and more commonly used as a long-term marker of inflammation. It is suspected to be a cardiovascular risk factor and to be correlated with endothelial dysfunction and subclinical atherosclerosis [19]. As an example, the authors of a meta-analysis concluded, that chronic use of phosphodiesterase inhibitors in in type 2 diabetes mellitus patients has a beneficial effect on endothelial function conceivably by a reduction of IL-6 serum levels [20]. For IL-8, a study by Shetelig et al. found that high levels were associated with larger infarct size and lower LV ejection fraction in STEMI patients [21]. In a study on IL-10, Jung et al. claimed, that IL-10 improves cardiac remodeling after AMI by stimulating M2 macrophage polarization and fibroblast activation [22]. The authors of another study concluded, that IL-10 inhibits inflammation and attenuates left ventricular remodeling after myocardial infarction via activation of STAT3 and suppression of HuR [23]. Further studies suggested a multitude of other biochemical pathways that might be involved in the complex relationship between these three cytokines and myocardial infarction [24‐32].

In the present study, two representatives of the fibroblast growth factor family, FDF-21 and FDF-23, were highly associated with 28-day mortality after AMI. Prior publications on FGF-23 suggested an association between elevated FGF-23 levels and cardiovascular events in general and mortality in CAD in particular [33‐37]. Though, the authors of a meta-analysis on this topic concluded, that the association might be non-causal [38]. In a rat model, Andrukhova et al. found that the induction of myocardial infarction in rats led to an elevation of FGF-23 levels [39], which strongly indicates a general association between AMI and FGF-23. Though, results of their study might rather suggest that AMI causes an elevation of FGF-23 levels and not the other way round. The results of the presents study would further indicate, that FGF-23 elevations are not only associated with the event itself, but also with adverse outcome after AMI.

For FGF-21, several prior studies suggested, that FGF-21 inhibits inflammation and fibrosis after AMI and therefore might improve cardiac remodeling [40‐42]. A review proposed FGF-21 as a new cardiomyokine which is crucial for maintaining cardiac function and has positive effects on the heart in the context of pathological conditions [43]. Therefore, we should have expected to find an inverse correlation between FGF-21 and mortality, which we didn´t. One possible explanation might be that very high levels of FGF-21 indicate great myocardial damage with a pronounced counter-regulation. Therefore despite having a protective effect, high FGF-21 levels are associated with higher mortality since they represent greater myocardial damage. This hypothesis is supported by a small study from China, which revealed significantly higher FGF-21 levels in AMI patients compared to non-AMI control patients, indicating that high FGF-21 levels are independently associated with AMI [44].

Another four biomarkers, i.e., MCP-1, CST5, ST1A1 and 4E-BP1, were strongly associated with 28-day mortality and consequently were included in the biomarker score. MCP-1 is a chemokine that regulates migration and infiltration of monocytes/macrophages [45]. A review article on cardiac repair after myocardial infarction reported, that early recruitment of pro-inflammatory monocytes is mediated through activation of the MCP-1/CCR2 axis and that in mice, MCP-1 inhibition exhibits a reduced infarct size and monocyte infiltration [46]. Next to a reduction of IL-6 serum levels as described above, it has been reported that phosphodiesterase-5 inhibitors like sildenafil also reduce MCP-1 levels in men with diabetes [47] and affect circulating monocytes and tissue inflammatory cell infiltration [48]. Another study including 87 AMI patients and 82 controls found higher levels of MCP-1 in the AMI group compared to a control group [49]. Additionally, they reported, that the highest MCP-1 levels were observed in patients with poor prognosis, which agrees with results of the present study.

For the remaining biomarkers CST5 (Cystatin D), ST1A1 and 4E-BP1 we couldn´t find prior studies that investigated a potential association between these biomarkers and AMI or AMI-related mortality, so we cannot compare our results to previously reported results. Further studies are needed to confirm the relationship of these newly detected protein biomarkers with adverse outcome after STEMI.

It could be conjectured, that the association between the nine biomarkers and short-term mortality after STEMI might be driven by the delay between symptom onset and PCI: patients with longer therapeutic delay have longer ischemic times which increases the plasma levels of the biomarkers at cardiac catheterization and at the same time adversely affects the outcome after AMI. Therefore we calculated the multivariable adjusted logistic regression models as described in the results section but added prehospital delay as an additional covariable (see Fig. 1, supplementary material). Although this influences the results of the model to a limited extent, it shows that therapeutic delay does not play a superior role in the association between the investigated biomarkers and short-term mortality.

Another important issue to consider is the well-known interconnection of pathological inflammatory processes with obesity and obesity-related diseases such as diabetes mellitus type 2. A general example for this association is an obesity-related increased susceptibility to infectious diseases [50], which also became visible in the context of the Covid-19 pandemic. But there is also increasing evidence that the link between obesity and inflammation plays a major role in metabolic and cardiovascular diseases [51‐53]. For instance, in a review paper Tarsitano et al. discussed the specific relation between epicardial adipose tissue and its pro-inflammatory role in several cardiovascular diseases [54]. It has moreover been reported, that specific cardio-protective agents like Phosphodiesterase-5 inhibitors can positively affect the course of the diseases by modifying underlying inflammatory processes [55, 56]. Against this background, one could have assumed that obesity in terms of an elevated BMI might be strongly involved the relationship between the inflammatory biomarkers and short-term mortality after AMI. Nevertheless, in the present study we found almost no differences for BMI between the groups and consequently no evidence indicating and substantial importance of BMI/obesity for the reported associations.

We found that the predictive ability of our calculated biomarker score is comparable to that of the well-established GRACE score. With AUC values of 0.7859 and 0.7961 both scores showed proper and comparable discrimination between cases (patients who died within 28 days) and controls. When both scores were combined to a total score, the predictive ability could be further increased (AUC of 0.8305) and according to cNRI and IDI the combined score added predictive ability to the GRACE score. Considering the results of the multivariable adjusted logistic regression models, it can be concluded that these nine markers are independently associated with short-term outcome after AMI.

These results may have two major implications. First, these markers could be measured in order to estimate a patient’s risk of adverse outcome, which then could be considered by physicians in therapy decisions. Currently, however, these markers are very rarely determined in the medical context and their measurement is time consuming and expensive, which makes them unfavorable candidates for early risk stratification unless a method for easy determination (e.g., an ELISA test) will be established.

Second, the markers found to have good predictive ability for short-term outcome might not only be associated with 28-day mortality but actually be causally related to mortality after AMI. If this would be the case, tailored treatment and pharmacological interventions might improve the outcome in AMI patients. Foremost, drugs that block the biological marker itself or that act on the corresponding receptors should be considered as a promising approach. Hartman et al. for instance gave a translational overview of anticytokine therapy in cardiovascular disease and after AMI [57]. They concluded, that promising results were mainly seen in experimental studies and in smaller clinical studies, but only one larger RCT showed positive results on outcome so far. But from our point of view the other biomarkers identified in this study would also deserve to be further investigated with regards to potential benefits of medical therapies. Nevertheless, to this day there is a lack of studies and evidence whether such new pharmacological approaches would indeed favorably affect the underlying pathophysiology of CAD and whether new drugs would improve outcome after AMI. The results reported in this study provide new starting points towards drug development and for future personalized treatment of AMI patients. Researchers in this field should strongly feel encouraged to further explore the possibilities that are revealed by this study.

This study is characterized by some strengths. First, this study is based on patients from the population-based myocardial infarction registry Augsburg with consecutive enrollment, which minimizes the effect of selection bias. Blood samples were uniformly taken immediately before the PCI intervention, guaranteeing highest consistency in blood sampling. Moreover, for every case there was a large number of additional information which we used for proper adjustment in the logistic regression models and allowed us to imitate the established GRACE score to compare predictive abilities.

Nevertheless, there are also some limitations. No validation cohorts from other registers are available for our analyses, preventing validation of the associations—in particular for the newly identified protein markers—found in this study. Moreover, as this study is based solely on observational data, we cannot draw any conclusions about causality (including the possibility of reverse causality). Even though the 92 markers measured in this study cover a broad spectrum of important and promising inflammatory markers, there are more inflammatory markers and plasma proteins that may have been of great importance in this context. Additionally, we might not have considered all relevant confounders. Furthermore, as this study included only STEMI patients with age between 25 and 84 years, the results may not be generalized to all age groups or ethnic groups as well as to Non-ST-elevation ACS events. Finally, as majority of the study cohort was male, it is not entirely clear whether the results can also be applied to female AMI patients. Prior studies have shown important differences in pathophysiology and response to treatment for both sexes in cardiovascular diseases [58, 59]; though the number of female patients in this study is insufficient for a valid subgroup analysis to address this limitation.