Developing A Predictive Model For Cancer Clinical Trial Accrual.

6557 Background: With increasing cost to conduct clinical trials, it is imperative to select trials rationally for local activation. We sought to create a predictive model with an outcome of anticipated accrual that can be used when considering a prospective clinical trial. Methods: This retrospective cohort study used 5.8 years of registry data to predict clinical trial accrual at our center. A negative binomial regression model was employed using variables known pre-study. These were abstracted from the OnCore clinical trial management system (Forte Systems, Madison, WI) and clinicaltrials.gov. Statistical significance was set a priori to 0.05. Normality and collinearity of independent variables and model fit were assessed. Variables modified an effect if the beta coefficient of the other independent variable changed by u003e10%. Accrual was predicted for the studies used to build the model. Results: The model included 207 trials with complete information. Mean accrual was 7.3 per trial (±18.4) and 55 (26.6...