Abstract
The aim of our observational study was to derive a small set out of 92 repeatedly measured biomarkers with optimal predictive capacity for adverse clinical events in heart failure, which could be used for dynamic, individual risk assessment in clinical practice. In 250 chronic HFrEF (CHF) patients, we collected trimonthly blood samples during a median of 2.2 years. We selected 537 samples for repeated measurement of 92 biomarkers with the Cardiovascular Panel III (Olink Proteomics AB). We applied Least Absolute Shrinkage and Selection Operator (LASSO) penalization to select the optimal set of predictors of the primary endpoint (PE). The association between repeatedly measured levels of selected biomarkers and the PE was evaluated by multivariable joint models (mvJM) with stratified fivefold cross validation of the area under the curve (cvAUC). The PE occurred in 66(27%) patients. The optimal set of biomarkers selected by LASSO included 9 proteins: NT-proBNP, ST2, vWF, FABP4, IGFBP-1, PAI-1, PON-3, transferrin receptor protein-1, and chitotriosidase-1, that yielded a cvAUC of 0.88, outperforming the discriminative ability of models consisting of standard biomarkers (NT-proBNP, hs-TnT, eGFR clinically adjusted) - 0.82 and performing equally well as an extended literature-based set of acknowledged biomarkers (NT-proBNP, hs-TnT, hs-CRP, GDF-15, ST2, PAI-1, Galectin 3) - 0.88. Nine out of 92 serially measured circulating proteins provided a multivariable model for adverse clinical events in CHF patients with high discriminative ability. These proteins reflect wall stress, remodelling, endothelial dysfunction, iron deficiency, haemostasis/fibrinolysis and innate immunity activation. A panel containing these proteins could contribute to dynamic, personalized risk assessment.Clinical Trial Registration: 10/05/2013 https://clinicaltrials.gov/ct2/show/NCT01851538?term=nCT01851538&draw=2&rank=1 .
