Formulating tightest bounds on causal effects in studies with unmeasured confounders