This issue of the British Journal of Clinical Pharmacology includes the first of a series of three papers concerning the opportunities and dangers presented by making interim analyses of accumulating clinical trial data. The term ‘interim analysis’ is used to describe an evaluation of the current data from an ongoing trial, in which the primary research question is addressed, and which has the potential for modifying the conduct of the study. The three papers each concern a different phase of clinical testing as the objectives of the various phases are quite distinct so that different methods are appropriate. The phases are taken in reverse order, because interim analysis methods have been most completely developed and applied to phase III, and are least familiar in phase I. The emphasis in the series is on opportunities, since the use of accumulating data to modify an investigation is a natural part of scientific enquiry, and because statistical methodology is now available to overcome the most serious dangers of misinterpretation.
The first paper is entitled ‘Interim analyses and sequential designs in phase III studies’. Here the potential for reducing trial duration and for avoiding allocation of patients to an inferior treatment is greatest, while the need to preserve the integrity of a rigorous statistical analysis is strongest. Methodology allowing the simultaneous achievement of both goals is described, and examples of completed trials which have used it are given. The second paper is entitled ‘Stopping rules for phase II studies’. In this context ethical and organizational needs for interim analysis are even greater, and numerous approaches have been proposed in the literature. These are outlined and compared, and examples are given of the sort of procedures that have been used in practice. The last paper is ‘Learning from previous responses in phase I dose-escalation studies’. Phase I trials conducted to identify the appropriate dose for further study are naturally conducted as a series of interim analyses. It is necessary to learn how one dose has affected subjects before administering a higher dose. However, formal statistical methods are seldom used in this process outside the context of cancer research. The paper will review techniques, such as the continual reassessment method, that have been implemented in oncology, and describe how similar approaches could be used over a wider range of therapeutic areas, including healthy volunteer studies.
The intention of this series of papers is to promote the use of interim analysis techniques in a way which does not compromise the validity of the analysis of a study, and to encourage collaboration between clinicians and statisticians in their implementation.