In personalized clinical trials, continuous biomarkers are often used to guide treatment assignment decisions by identifying patients more likely to benefit from an experimental therapy. A common approach is to discretize these biomarkers via a cut-off/threshold given by a data-driven quantile, usually the median, in order to define the subgroups. Although easy to apply, this strategy can lead to questionable decisions from both an ethical and practical perspective. In this work, we take into account the design-based framework introduced by Baldi Antognini et al.[1] as a starting point to reflect on what happens when inappropriate thresholds are used to guide treatment assignments in comparative trials with binary outcomes and a continuous covariate. Our main focus is on the ethical dimension: we highlight how arbitrary cut-offs can lead to unfair and potentially harmful treatment assignment decisions. We also discuss some economic consequences, such as wasted resources due to misclassification. We simulate several biomarker distributions to compare the optimal, model‐based threshold of [1] with the common median‐split rule, quantifying ethical and economic impacts such as allocation to suboptimal treatments. FDA warnings note that poor cutoff selection jeopardizes trial integrity and patient welfare [2]. This issue is timely given the expanding role of biomarkers in early‐phase trials and regulatory decisions. Inadequate thresholds can compromise trial interpretation, restrict patient access to promising therapies, and undermine fairness and transparency. We advocate for model‐based threshold rules grounded in clinical insight and statistical rigor to enhance transparency for researchers, clinicians, and regulators.