The proliferation of biomarkers has raised concerns regarding the possibility for clinical judgment to be improperly removed from clinician's jurisdiction and included in laboratory tests. To evaluate the ways in which the diffusion of biomarkers questions the autonomy of clinicians, we consider the case of chemotherapy prescription to women with early stage breast cancer and a good prognosis. Drawing on a qualitative study of clinicians working in a diversity of institutional contexts, we follow three biomarkers available to guide this routinely made decision. We show that, biomarkers able to reduce all the uncertainties associated with, what we analyse as an uncomfortable decision, are sought more than dreaded by clinicians. If such ideal tools are unavailable, the fact is well acknowledged by the profession. Rather than precluding their usage, the imperfection of existing biomarkers is controlled by the profession, through their integration as additional tools in the decision process. The fact that the biomarkers are recognized as imperfect biomedical entities reinforces the importance of local material, organizational and financial constraints over that of international science, technology and clinical data, in their diffusion. The regulation of the uncertainties associated with these imperfections is organized at the professional level. Through an important work, relying on guidelines and enforced in collective bodies, the series of heterogeneous bioclinical evidences available are articulated. Biomarkers tend to be subordinated to the clinic. While maintaining the professional autonomy, the process also strengthens the internal professional hierarchy. When the most expert clinicians manage to inhabit a space for clinical autonomy, the nonexpert are torn between stronger professional rules and patient preferences. In this alliance between biomarkers and experts, their clinical autonomy tends to be the price for the professional autonomy.