Development of a mathematical model for thrombosis risk prediction using serum biomarkers

Abstract

Introduction. Thrombosis is a frequently underdiagnosed condition associated with high mortality in neglected cases. Many factors, including geoheliophysical and biochemical ones, are responsible for thrombosis modulation. Routine investigations may sometimes be inconsistent and, thus, unreliable in a clinical setting.

Material and methods. Data were collected from patients treated in the Department of Vascular Surgery at the ‘Timofei Moșneaga’ Republican Clinical Hospital, Chișinău, Republic of Moldova. A total of 1,865 patients were initially included in the study. After applying rigorous inclusion and exclusion criteria, 263 eligible patients were identified, and their complete blood counts and biochemical reports were retrospectively analyzed.

Results. The analysis revealed increased mean values for absolute polymorphonuclear neutrophils, absolute monocytes, erythrocyte sedimentation rate (ESR), and glucose. The median values of these indicators, except for absolute polymorphonuclear neutrophils and ESR in female patients, were also elevated above normal ranges. Significant Pearson and Spearman correlations were identified among the analyzed indicators, and a binary logistic regression model was constructed using the most statistically significant variables.

Discussion. Usual mathematical models that outline thrombosis consider deep vein thrombosis without a sustainable arterial assessment. The sensitivity of our model is lower than that of the D-dimer, while the specificity is almost the same. Platelets and clotting tests are well-known, reliable indicators; however, novel contemporary augmentations to these may, in turn, increase the predictive capability of our model if applied. This study has its limitations due to the lack of variance in the variance inflation factors (VIF), preventing the evaluation of multicollinearity among the included biomarkers.

Conclusions. The mathematical model developed in this study shows potential for further clinical application; however, additional research, validation, and the incorporation of non-biochemical indicators may be necessary to enhance its predictive accuracy.