Abstract
Efficacy of thromboelastography (TEG)-derived parameters for the prediction of acute thromboembolism (AT) in patients with non-steroidal anti-inflammatory drug (NSAID)-induced gastrointestinal bleeding (GIB) remains to be determined. A retrospective propensity score matching (PSM) study was performed to evaluate this efficacy. Patients with NSAID-induced GIB (98 with AT; 830 without AT) were matched for age, sex and history of cardiovascular and cerebrovascular diseases using PSM. Multivariate logistic regression was used to determine the efficacy of TEG-derived predictors of AT. Mean Decrease Gini (MDG) coefficients were used to rank the importance of the variables from random forest algorithm results. Univariate analysis indicated that the following indexes were significantly different between the two groups: Reaction time (R value), coagulation forming time, solidification angle, maximum amplitude (MA), coagulation index (CI), hemoglobin levels, D-dimer levels, platelet aggregation test (pAgt) results, fibrinogen levels and Acute Physiology and Chronic Health Evaluation II score (all P<0.001). Multifactor logistic regression analysis indicated that the R value (P=0.010), solidification angle (P=0.004), MA (P=0.038), D-dimer levels (P=0.012) and pAgt results (P=0.015) were independent predictors of AT in patients with NSAID-induced GIB, achieving an area under the curve of 0.999 in receiver operating characteristic curve analyses. The five most important parameters according to the MDG scores (MDGS) were: Solidification angle (MDGS=58.14), R value (MDGS=20.42), pAgt results (MDGS=15.61), D-dimer levels (MDGS=12.78) and CI (MDGS=12.61). The results of the present study indicated that TEG-derived parameters including the R value, solidification angle, MA and CI, as well as D-dimer levels and pAgt score were significant predictors of AT in patients with NSAID-induced GIB.