Liver parameters as part of a non-invasive model for prediction of all-cause mortality after myocardial infarction

Figures

• 

  Figure 1. Recruitment and exclusion flow of patient cohort. Initially data of 1242 patients were screened retrospectively. All patients not fulfilling study criteria or without available follow-up were excluded. The remaining 569 patients were grouped according to survival

• 

  Figure 2. Unadjusted survival curves and cumulative incidence of MACE. Incidence of cardiac death (A) is given over a follow-up time of 70 months. Only individuals with a known outcome are depicted. Deceased patients for whom time of death was not known were censored at 70 months. Cumulative incidences of major adverse cardiac events (MACE) are depicted for unstable angina (UA; B), restenosis (C), and reinfarction (D). Only individuals with known outcome were included in this analysis (351 for UA and reinfarction; 350 for restenosis). Patients for whom time of the MACE occurrence was not known were censored at 62 months (UA, restenosis) or 50 months (reinfarction)

• 

  Figure 3. Ensemble feature selection importance analysis. The bars show the cumulative importance for the different parameters with respect to separation of death and survival. Higher numbers indicate greater importance for the classification. Partitions in the bars show contributions of single features of the ensemble selection approach to the total importance P_cor – Pearson correlation, S_cor – Spearman correlation, LogReg – logistic regression, ER_RF – error rate of random forest, Gini_RF – Gini impurity in random forest, AUC_CF – area under the curve for conditional random forest, ER_CF – error rate of conditional random forest, eGFR – estimated glomerular filtration rate, Mygl – myoglobin, Trpn – troponin, Crtn – creatinine, CRP – C-reactive protein, Blrb – bilirubin, Gndr – gender.

• 

  Figure 4. Performance of prediction model for prediction of cardiac death. (A) On the Y-axis the sensitivity and on the X-axis the specificity is shown. The ROC curve is shown as a solid line. The AUC of the prediction model is 77.6% (95% CI: 72.5–82.6%). The confidence interval is shown as a dashed line. The dotted line marks the performance of random guessing. (B) Kaplan-Meier analysis for C-reactive protein (CRP) as the most important parameter for prediction of death after acute myocardial infarction (AMI). Only patients with known time of death (n = 76) were included in the analysis and grouped according to low and high CRP (cutoff at 4.2 mg/dl). Patients with high CRP had significantly shorter survival time after AMI