Estimating the causal effect of compliance on binary outcome in randomized controlled trials