A Predictive Model for Chemotherapy-Associated Thrombosis Based on Findings from a Prospective Outpatient Registry.

Background: Venous thromboembolism (VTE) is a significant complication of cancer and particularly of newer anti-cancer therapies. The risk of VTE is estimated to be 0.04%/month (0.5%/year) in cancer patients. We conducted a prospective analysis to determine the frequency and risk factors for symptomatic VTE in cancer patients actively receiving chemotherapy.

Methods: The Awareness of Neutropenia in Cancer (ANC) Study Group is currently enrolling patients with breast, lung, colon, ovary and other cancers, at 137 sites nationwide, with an accrual goal of 4,500 patients. Patients are registered at the time of initiation of a new chemotherapy regimen, and followed for the duration of 4 cycles. The association of VTE with clinical variables was studied in univariate analysis and in a multiple logistic regression model. No significant first-order interactions were observed between the variables in this model. A risk score for VTE was estimated for each subject from the weighted sum of model predictor variables from the multivariate model.

Results: A total of 2,151 patients treated with at least one cycle of chemotherapy were available for analysis. Pulmonary embolism occurred in 11, and deep venous thrombosis in 35, for a VTE incidence of 2.14 % over median follow-up of 2.5 months (0.85%/month). Incidence of first VTE event was greatest in cycles 1 (0.93%/cycle) and 2 (0.88%/cycle), and declined in cycle 3 (0.53%/cycle). Incidence varied significantly by site of disease (p=0.01) with highest rates in patients with upper gastrointestinal (2.4%/month) and lung cancers (1.2%/month), and lymphomas (1.2%/month). Other clinical variables associated with development of VTE in univariate analysis included baseline platelet count ≥350,000, an ECOG PS ≥ 2, and baseline use of erythropoietin or colony-stimulating factors. Patients with VTE reported a chemotherapy delay of ≥ 7 days in 40% compared to 23.6% in patients without VTE (OR 2.15, p=0.01). Based on the results of the multivariate model (c statistic=0.73[.65,.81]) the population was divided into low, intermediate and high-risk based on tertiles of risk scores. Patients deemed high-risk had a VTE risk of 1.5%/month, which was a six-fold increase compared to those deemed low-risk (see Table).

Conclusion: Symptomatic VTE is much more frequent in cancer patients on chemotherapy than previously estimated, even in patients with hematologic malignancies. VTE is associated with a nearly two-fold risk of significant delay in chemotherapy, which could impact on cancer outcomes. This is the first prospectively generated predictive model for chemotherapy-associated thrombosis, and can be used to define a high-risk study population for trials of thromboprophylaxis.

Incidence of Chemotherapy-Associated VTE By Risk Categories