Null results in clinical trials: the need for a decision-theory approach.

A framework is developed to take explicitly into account the conflicting demands of ethics and scientific rigor in the design of clinical trials. The framework recognizes the part played by the clinical-scientific community in the weighing of a new result provided by a clinical trial. To illustrate the usefulness of the framework, a value system is adopted which gives relatively high weight to ethical considerations. The analysis based on this value system reveals some limitations of the present clinical-trials mechanism, especially if success is defined exclusively in terms of cure, and other dimensions of the health system, such as explanatory, care, cost or prevention variables are neglected. On the basis of this analysis, it is suggested that: (a) If randomized clinical trials are to be ethically acceptable, they will necessarily yield a large proportion of null results. (b) Positive results from ethically acceptable clinical trials would be expected to have less impact than null results; unless this is the case, there will be a tendency to encourage false hopes. (c) Trials need not yield entirely null results, provided that attention is not focused exclusively on a single outcome variable. A trial of chemotherapy for acute myeloid leukaemia in adults is used to illustrate the need for new approaches to the planning and design of clinical trials.