[Problems of clinical trials as viewed in the perspective of decision theory]

In this paper, the current clinical trial based on the Neyman-Pearson's principle of hypothesis testing is reviewed in the perspective of decision theory, its shortcomings are revealed and alternative methods proposed. In the current trial design, the alpha error is arbitrarily fixed at a level (e.g., 0.05), even lower than the beta error level (0.1 or 0.2). This results in failure with smaller trial size to identify the better treatment, dose-patient interaction, or optimal patient stratum for treatment. If on the other hand, the most harmless alpha error is ignored, the total number of patients expected to be cured can be substantially increased due to earlier identification of the better treatment. In case treatment, A has some disadvantages compared with B in terms of cost, toxicity etc., then the null hypothesis should be changed to "the cure rate of A is greater than that of B by delta", an adjustment for the disadvantages, which can be estimated by utility analysis. The loss to the society and target population may be further increased if improper regimen is provided including omission of the best treatment arm, suboptimal treatment intensity, selection of improper patient stratum, failure to individualize treatment plan (inflexible regimen). In order to maximize the benefit of our patients and society, radical changes of the current clinical trial design is needed using decision analytical approach.