Sample size considerations in active-control non-inferiority trials with binary data based on the odds ratio