PB1135 - Machine Learning for Predicting Thrombosis in Cancer Patients: A Systematic Review

Background: Khorana score (KS) is an established risk assessment model for predicting thrombosis in cancer patients. However, it ignores several risk factors and has poor predictability in some cancer types. Machine Learning (ML) is a novel technique being used for the prediction of several diseases, including cancer-associated thrombosis, when trained on specific diagnostic modalities. Consolidating the literature on the use of ML for the prediction of cancer-associated thrombosis is necessary to understand and improve its predictability relative to KS.

Aims: This systematic review aims to evaluate the current use and performance of ML algorithms to predict thrombosis in cancer patients.

Methods: This study was conducted per PRISMA guidelines. Databases Medline, EMBASE, Cochrane, and ClinicalTrials.gov were searched from inception to September 15 2023, for studies evaluating the use of ML models for the prediction of thrombosis in cancer patients. Keywords 'machine learning', 'artificial intelligence', ‘thrombosis’, and ‘cancer' were used. Studies examined adult cancer patients using any ML model were included. Two independent blinded reviewers conducted study selection and data extraction.

Results: 300 citations were screened, of which 29 studies underwent a full-text review, and ultimately, 8 studies with 22,893 patients were included (Fig.1). Sample sizes ranged from 348 to 16,407 patients. Thrombosis was characterized as venous thromboembolism (n=6) or peripherally inserted central catheter thrombosis (n=2). The types of cancer included breast, gastric, colorectal, bladder, lung, esophageal, pancreatic, biliary, prostate, ovarian, genitourinary, head-neck, and sarcoma. All studies reported outcomes on the machine learning’s predictive capacity (Table 1). The XGBoost appears to be the best-performing model, and several models outperform KS in their respective datasets.

Conclusion(s): Certain ML models allow for accurate prediction of thrombosis in cancer patients and can outperform KS. Limited datasets and variability in algorithm metric scores warrant additional studies to supplement the current ones.