Abstract Number: PB1054
Meeting: ISTH 2021 Congress
Background: The antiphospholipid syndrome (APS) is characterized by the presence of antiphospholipid antibodies (aPL) predominantly directed against β2-glycoprotein I. APS is associated with an increased risk of thrombosis and pregnancy morbidity. Diagnosing APS is difficult because most patients are already on anticoagulation when tested for aPL and anticoagulant treatment interferes with aPL assays. Nevertheless, the aPL profile defines patient management, making aPL testing warranted during anticoagulant therapy.
Aims: We developed a neural net (NN) that diagnoses APS in a cohort of anticoagulated patients and controls based on thrombin generation (TG) and thrombin dynamics.
Methods: A NN was developed using TG data obtained with the CAT method in 48 APS patients anticoagulated with vitamin K antagonists and 64 anticoagulated controls. Input parameters were lag time, peak, ETP, time-to-peak, velocity index, total prothrombin conversion, maximum prothrombin conversion rate, thrombin-antithrombin, thrombin-α2-macroglobulin, and the thrombomodulin effect. Five NNs were developed and the most accurate NN was selected and clinically validation in the validation cohort of 311 APS patients and controls. The golden standard for APS diagnosis were the ISTH-SSC guidelines.
Results: In the derivation cohort, the NN classifies APS patients under anticoagulant treatment with a sensitivity of 92% and a specificity of 95%, (ROCAUC = 0.9805; 0.9542-1.000; p<0.0001). In the validation cohort, the NN was clinically validated in 33 APS patients and 278 controls, including anticoagulated controls (n=62), thrombosis patients (n=38), auto-immune disease patients (n=49), patients visiting the hospital for other indications (n=92), and normal controls (n=37; Figure 1). The sensitivity of the NN was 85%. The specificity of the NN was 93% in the whole validation cohort and ranged from 100% in normal controls to 76% in thrombosis controls (Table 1).
Conclusions: We developed a NN that accurately classifies APS under anticoagulant treatment. This NN could be an alternative for the LAC test which is affected by anticoagulation.
To cite this abstract in AMA style:
de Laat-Kremers R, Wahl D, Zuily S, Ninivaggi M, Chayoua W, Regnault V, Musial J, de Groot P, Devreese K, de Laat B. Artificial Intelligence Classifies APS in Anticoagulated Patients Based on Thrombin Generation [abstract]. Res Pract Thromb Haemost. 2021; 5 (Suppl 2). https://abstracts.isth.org/abstract/artificial-intelligence-classifies-aps-in-anticoagulated-patients-based-on-thrombin-generation/. Accessed April 19, 2024.« Back to ISTH 2021 Congress
ISTH Congress Abstracts - https://abstracts.isth.org/abstract/artificial-intelligence-classifies-aps-in-anticoagulated-patients-based-on-thrombin-generation/