Validation of venous thromboembolism predictive model in hematologic malignancies

Area of Science:

• Hematology
• Oncology
• Vascular Medicine

Background:

• Existing venous thromboembolism (VTE) risk stratification scores inadequately represent patients with hematologic malignancies (HM).
• There is a need for validated predictive tools to assess VTE risk specifically in hospitalized HM patients.

Purpose of the Study:

• To develop and validate a logistic regression model for predicting VTE risk in hospitalized HM patients.
• To assess the accuracy and discriminative ability of the developed VTE predictive model through internal and external validation.

Main Methods:

• A prospective case-control study involved 496 hospitalized HM patients (December 2010-2020).
• A logistic regression model was developed using data from 285 patients, incorporating factors like hypercholesterolemia, tumoral activity, thrombogenic drugs, diabetes mellitus, and immobilization.
• Internal validation used bootstrap analysis, and external validation was performed on a separate cohort of 211 HM patients.

Main Results:

• The VTE predictive model demonstrated strong performance in internal validation with an AUC of 0.838 and accuracy of 80.1%.
• External validation showed high accuracy (82.9%), an AUC of 0.900, and favorable predictive values (NPV 89.7%, PPV 67.7%).
• The model exhibited good sensitivity (74.6%) and specificity (86.2%) in the external validation cohort.

Conclusions:

• The developed VTE predictive model is a reproducible and simple tool for assessing VTE risk in hospitalized HM patients.
• The model exhibits good accuracy and discrimination, supporting its clinical utility.
• This validated tool can aid in proactive VTE risk management for hematologic malignancy patients.