Prescribing before diagnosis is settled is common in clinical medicine, but empirical treatment becomes difficult to interpret when the therapeutic objective, expected time course and exit conditions remain implicit. Diagnostic uncertainty is not itself an indication for treatment. A therapeutic trial is defensible only when exposure is clinically justified despite unresolved diagnosis, the treatment has a plausible relation to a defined target problem, and the clinician can specify what observations would alter the decision to continue, modify or stop. This commentary argues that empirical prescribing should be designed prospectively as a bounded clinical test. The central requirement is not formal N-of-1 experimentation in every patient, but enough prospective structure to distinguish treatment utility from diagnostic confirmation and to reduce therapeutic drift. Apparent improvement may reflect pharmacological benefit, spontaneous fluctuation, regression toward a usual state, expectancy effects or concurrent care; nonresponse can be equally ambiguous. Reassessment timing must follow mechanism, risk and the information likely to emerge during treatment. Explicit stopping rules complete the logic by making continuation an active decision. This approach narrows the legitimate role of empirical treatment while improving the interpretability and reversibility of prescribing under uncertainty.
