Event projection is important for clinical trials with time to events as the primary endpoints. The sponsors estimate the study duration and time for interim analysis at the design stage based on the best assumptions of event rates and subject accrual. During study conduct, real-time monitoring and prediction of enrollment, events and loss of follow up help the sponsors correct the inappropriate assumptions and plan the future activities. Though many parametric and nonparametric methods have been proposed for event projection, most of them assume the treatment group information is known. Donovan et al (2006) proposed a mixture model for the situations when treatment arm is masked and used the Bayesian prediction distribution for the projection. We propose a simpler version to avoid the computation burden and also incorporate the all available information to improve the prediction. The comparison is made through simulations to demonstrate our model's performance characteristics.