Plasma chemokines indicate enhanced bleeding in patients with chronic coronary syndrome undergoing percutaneous coronary stenting

Figures

• 

  Fig. 1. Plasma levels of chemotactic cytokines are associated with a substantial modulation of ex vivo platelet function. A Pattern analysis of circulating chemokines (t48 h) associated with collagen-mediated platelet aggregation. Bars represent Spearman correlation coefficients and are coloured according to the direction of the relationship. Asterisks represent significant (p < 0.05) correlations. B Correlation analysis of adenosine diphosphate (ADP) induced platelet aggregation and plasma chemokine concentration at t48 h. C sGPVI concentration in patients with CAD is associated with a shift in the chemokine profile (t48 h). Compounds are highlighted according to the referring correlation coefficient. D Correlation matrix synopsizing correlations of chemokines (t48 h), study drug treatment, and sGPVI concentrations alongside ex vivo platelet functional assays. Spearman’s ρ is coloured and dots are plotted proportional to correlation coefficients. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

• 

  Fig. 2. CAD patients with adverse bleeding events exhibit a substantial deficiency of chemotactic eotaxin. A Bar chart subsumes comparison of chemokines between patients with relevant bleeding to those without adverse events at baseline (t0 h). There were no observations of an altered chemokine signature prior PCI and administration of antiplatelet therapy or study drug treatment. B Sankey plot depicting proportions of adverse bleeding events between individual study drug treatment groups. Thus, treatment with GPVI antagonist revacept was not associated with bleeding incidences during the follow-up period. C In the overall cohort bar charts compare chemokine concentrations after PCI and allocation of study drug (t48 h) in line of bleeding events. *p < 0.05. D Postinterventional eotaxin concentration (t48 h) was significantly (p < 0.05) reduced in patients with relevant bleeding (BARC 2–5) compared to patients without adverse events regardless of whether they received revacept or placebo E Bar chart exhibiting the percentage change of distinct chemokines before and after PCI and study drug administration. *p < 0.05. F Plasmatic eotaxin is critically (p < 0.05) decreased in patients with…

• 

  Fig. 3. The plasma chemokine profile is influenced by the antiplatelet treatment with revacept in patients with coronary artery disease. A Hierarchical clustering analysis of the analysed chemokines (t48 h) according to the referring treatment group. Z-scores of median concentrations were coloured according to up-regulation (red) or down-regulation (blue) of distinct mediators. B Bar chart showing the dose-dependent regulation of chemokines (t48 h) in revacept-treated patients in contrast to placebo. None of the chemokines differed significantly (p < 0.05) between treatment groups. C Collagen-dependent platelet aggregation was significantly (p < 0.05) decreased in patients with relevant bleeding after administration of Revacept when compared to patients receiving placebo (t48 h). D Ex vivo collagen-dependent platelet aggregation did not differ between treatment subgroups in patients without adverse events (t48 h)

• 

  Fig. 4. Machine learning identifies a chemokine signature in CAD associated with adverse bleeding events after PCI and antiplatelet therapy. A Methodological outline of the risk estimation employing XGBoost models of risk parameters including plasma chemokines (t48 h) in CAD patients of this study. B Most important variables predicting adverse bleeding events based on a machine learning algorithm (XGBoost). Included parameters are sorted according to relative importance of gain parameters and mostly comprise plasma chemokine levels. C Receiver operating characteristic (ROC) curves depicting the diagnostic accuracy of XGBoost model of the training cohort. Area under the curve (AUC = 0.93) of the cross-validated machine learning algorithm yields a high degree of accuracy in identifying patients with bleeding events. D Transformed likelihood formula of bleeding risk from XGBoost model was plotted against actual bleeding events. A substantial (p < 0.05) separation of patients with relevant bleeding from those without adverse bleeding demonstrates the accuracy of risk assessment integrating plasma chemokine levels. E Kaplan–Meier curves subsuming quartiles according to the individual bleeding…