Risk prediction model for thrombosis in leukemia patients: a systematic review.

Area of Science:

• Hematology
• Oncology
• Clinical Epidemiology

Background:

• Thrombosis is a major complication in leukemia patients, impacting treatment and survival.
• Existing risk prediction models for leukemia-associated thrombosis have uncertain quality and applicability.

Purpose of the Study:

• To systematically review and evaluate the methodological quality and validation of existing risk prediction models for thrombosis in leukemia patients.

Main Methods:

• Comprehensive literature search across nine databases up to August 4, 2025.
• Independent study selection, data extraction, and quality assessment using CHARMS and PROBAST tools.
• Included 14 studies with 16 prediction models, assessing discrimination (AUC/C-index) and validation methods.

Main Results:

• Models showed moderate to good apparent discrimination (AUC/C-index 0.641-0.917).
• Key predictors identified: CVC, prior thrombosis, D-dimer, platelets, WBC, ECOG score, treatment, comorbidities, leukemia type, hemostatic drugs, age.
• Most studies had high risk of bias; only one external validation was performed; significant concerns regarding applicability were noted.

Conclusions:

• Current thrombosis risk prediction models for leukemia patients are limited by methodological weaknesses and inadequate validation.
• Models should be considered exploratory and not for routine clinical use without prospective external validation.
• Future research needs prospective, multicenter studies with rigorous validation across diverse leukemia subtypes.