Causal inference under outcome-based sampling with monotonicity assumptions