A predictive model for venous thromboembolism in patients with germ-cell tumors receiving chemotherapy.

e15545 Background: To develop a prediction model for venous thromboembolism (VTE) in patients with germ-cell tumors receiving chemotherapy. Methods: We designed a multi-institutional study with a sample size estimation of 660 patients to develop a model with 5 explanatory covariates. Patients with germ-cell cancer treated between 2000-2014 were pooled from the Spanish Germ-Cell Cancer Group (Grupo Germinal)/Spanish Oncology Genitourinary Group(SOGUG) registry and another 3 referral hospitals. The primary endpoint was the ocurrence of VTE. We applied a logistic regression with Penalized Maximum Likelihood Estimations (PMLE). The c-index corrected for possible overfitting using the bootstrap method was calculated to validate the model. Results: We enrolled 683 consecutive patients from 17 hospitals. Platinum-based chemotherapy was initially used in all cases: BEP (86%), EP (11%) and others (3%). A total of 56 VTEs (8%) were observed. Explanatory covariates in the PMLE regression model were: IGCCCG risk scor...