Wouldn't you have to mathematically account for a 2 to 1 randomization to reach assuredness of being at the median before unblindnig? Realistically, the number could be less than 146 with a 99% or 98% level of confidence, but how much less? Maybe we could get a little help from our stats folks.