Abstract 3294
Background
Germ-cell cancer patients (GCC) treated with chemotherapy are at high risk of developing venous thromboembolic events (VTE). Several single parameters have been proposed for risk prediction, but no risk assessment model (RAM) is currently available.
Methods
All previously reported variables related with venous thrombosis in GCC were included. The association of these predictors with the development of VTEs was assessed using an Elastic Net penalized logistic regression model. The selection of the penalization factor lambda for the Elastic Net analysis was performed using 10-fold cross-validation. The adjusted model was validated using an external and independent cohort. As a measure of performance of the model the AUC (area under ROC curve) was used. The training dataset consisted of all consecutive chemotherapy-treated GCC patients recruited by 13 hospitals in the Spanish Germ-Cell Cancer Group (SGCCG) registry between 2004 and 2014. The validating dataset consisted of an external and independent cohort of patients treated at 4 independent hospitals.
Results
The training subset included 513 patients, with a VTE rate of 9% (N = 44). Our penalized logistic regression model included 4 predictors: Large retroperitoneal mass (N3), presence of liver, bone or brain metastasis (LBB), IGCCC poor prognosis (IGCCCPoor) and haemoglobin basal level (HBB), which were able to predict the probability of an event following the equation: P(event) = e − 2.33 +0.16*N3+0.27*LBB+0.19*ICCCCPoor-0.003*HBB/(1 + e − 2.33+0.16*N3+ 0.27*LBB+0.19*ICCCCPoor - 0.003*HBB). This model resulted in an area under the ROC curve of 0.83 (95% CI: 0.76-0.90). The validation subset included 325 patients with a VTE rate of 13% (N = 42). The area under the curve in the validation sample was 0.73 (95% CI, 0.65, 0.83).
Conclusions
We report a risk assessment model able to predict the probability of VTE in chemotherapy-treated GCC. This model may guide patient selection in studies of thromboprophylaxis in this population.
Clinical trial identification
Legal entity responsible for the study
Spanish Germ Cell Cancer Group
Funding
Spanish Germ Cell Cancer Group
Disclosure
All authors have declared no conflicts of interest.