Development of a clinical prediction tool for cancer-associated venous thromboembolism (CAT): the MD Anderson Cancer Center CAT model

• Published in: Supportive care in cancer 2020-08, Vol.28 (8), p.3755-3761
• Main Authors: Rojas-Hernandez, Cristhiam M., Tang, Victoria K., Sanchez-Petitto, Gabriela, Qiao, Wei, Richardson, Marsha, Escalante, Carmen
• Format: Article
• Language: English
• Subjects: Cancer; Chemotherapy; CT imaging; Erythropoietin; Health risk assessment; Hematology; Medicine; Medicine & Public Health; Metastasis; Nursing; Nursing Research; Oncology; Oncology, Experimental; Original Article; Pain Medicine; Rehabilitation Medicine; Thromboembolism; Thrombosis
• ISSN: 0941-4355; 1433-7339
• DOI: 10.1007/s00520-019-05150-z

Introduction Cancer-associated venous thromboembolism (CAT) is a major complication of malignancy. Our goal was to develop a prediction model for VTE that better represented to the population seen at large referral cancer centers. Materials and methods This study was nested in a prospective cohort study at the University of Texas MD Anderson Cancer Center that evaluated adult patients during outpatient cancer-staging computed tomography to estimate the prevalence of incidental VTE. Data from patients in whom incidental VTE was not found on initial CT were collected until 24 months ± 7 days from the study inclusion date to determine the occurrence of new VTE events. Demographics, clinical data, current cancer treatment information, and the use of erythropoietin stimulating agents (ESAs) along with hematologic variables were collected in all patients and analyzed to determine differences between those who developed VTE versus those who did not. All candidate variables with significance p value (≤ 0.1) under univariate analysis were considered to enter the final multivariate model. Results Data of 548 patients were analyzed. The presence of metastatic disease and the use of platinum-based chemotherapy were strongly associated with CAT occurrence. The use of ESAs and specific malignancies showed trends of association with CAT, while associations were not statistically significant.Those characteristics were utilized to develop a clinical prediction model for CAT readily available and effective (c-index = 0.74). Conclusion Our model is effective and easy to incorporate in busy clinical settings and it does not depend on esoteric or difficult-to-obtain laboratory testing. Future external validation studies may provide further evidence for the applicability of our results.