Abstract Body (Do not enter title and authors here): Background: End-stage heart failure patients undergoing left ventricular assist device (LVAD) implantation often face thrombotic complications (TC) such as pump thrombosis, venous thromboembolism, and ischemic stroke. This study used untargeted plasma metabolomics to identify potential biomarkers predictive of TC.

Methods: In an existing cohort of 100 LVAD patients, a subgroup of 25 patients was investigated. This exploratory study included 8 patients who developed TC within 6 months of LVAD implantation, as well as 17 patients who did not experience TC during the same period. Untargeted plasma metabolomics using LC-MS/MS was conducted on day 0 (pre-LVAD) and day 7 (post-LVAD). A comprehensive, web-based tool, MetaboAnalyst was used to train classifiers to distinguish TC vs No TC. Enrichment analysis was conducted using parent classes in the Human Metabolome Database (HMDB). The patients’ coagulation status was also monitored using the International Normalized Ratio (INR) every week for the first month, and regression analysis was used to test for correlation between the INR and the metabolites that predicted TC.

Results: TC occurred between days 35 and 140. Among 1473 metabolites, 35 differed between TC vs No TC in pre-LVAD plasma (Fig.1a) and 73 in post-LVAD plasma (Fig.1b). The post-LVAD classifiers had an area under the curve (AUC) of 0.704-0.741 (Fig.2a). Key markers included glucuronides and methylephedrine, which were increased in TC; and metabolites of tramadol, acetaminophen, and caffeine, which were decreased in TC (Fig.2b). Enrichment analysis showed differences in O-glucuronides, 6-oxopurines, and aralkylamines (p<0.05, Fig.3). In heatmap clustering, O-glucuronides formed a single cluster with elevated levels in patients who would develop TC and 6-oxopurines formed a small cluster with lower levels in TC, but aralkylamines were not grouped. Cinnamoyl glucuronide was correlated with subtherapeutic INR at week 3 (t=3.27, p<0.01), but in multiple logistic regression, cinnamoyl glucuronide predicted TC independent of INR.

Conclusion: Drug metabolites like caffeine, tramadol, and acetaminophen on day 7 influenced TC risk for up to 140 days. Increased glucuronide levels in TC patients were correlated with subtherapeutic INR, suggesting that overactive glucuronidases may be metabolizing warfarin and predisposing to thrombosis. Further metabolic studies may help to personalize thrombosis prevention in LVAD patients.