Treatment rules are decision functions designed to assign treatments so as to maximize individual-level outcomes — such as healing or survival. When these rules are learned from observational or experimental data, they may inherit biases if certain strata of the population are underrepresented, potentially leading to unfair treatment recommendations. In this work, we address this issue by defining fair treatment rules under a demographic parity (DP) constraint. We treat treatment allocation as a classification problem and enforce DP on the final binary decision by modifying a decision threshold. We define the optimal treatment rule under DP constraint problem and analyse the “cost of fairness” as well as excess risk and constraint violations in its implementation. Our results highlight how enforcing DP at the level of the treatment decision alters the resulting policy.