Clinical trial design has a strong history of statistical innovation to ensure evidence is sufficient to draw conclusions and influence clinical practice. Typically, trials control the size and power of a hypothesis test on a pre-specified primary outcome, selected using clinical expertise. However, relying on statistical significance has been criticized over the years as, among other issues, it has encouraged the use of trial outcomes with limited practical relevance and statistical and clinical significance may not coincide. This has reduced the impact of clinical research in practice.
Value of Information (VoI) methods are concepts from decision theory that are a potential to standard clinical trial design. VoI methods compute the value, measured in monetary or health units, of the evidence that would be collected in a proposed trial. The research value can be compared with its cost to target trials where the return on research investment is high. VoI goes beyond sample size and power considerations and, alongside clinical judgment, can be a comprehensive, principled method for
- prioritizing research areas,
- selecting trial outcomes,
- fixing trial sample size and
- evaluating competing trial designs.
Despite these advantages, the large computational cost of VoI analyses has been a significant barrier to their practical use. To overcome this, my recent research has developed and evaluated novel algorithms that allow these methods to be used. However, barriers to implementation remain, e.g., the lack of appropriate software, difficulties fully characterizing uncertainty and challenges evaluating non-standard trial designs.
Objective of Proposed Research
The long-term goal of this research program is to improve the use of VoI methods in clinical trials design through the development of relevant methodology, guidance and software.
The short-term objectives of this program are to:
- facilitate the use of novel VoI calculation methods through the development of a purpose-built software package.
- improve the accuracy of VoI methods in the face of structural uncertainty and immature evidence.
- augment VoI calculation methods to evaluate and design adaptive trials.
Common methods for trial design focus on statistical significance and ignore that decision-making about clinical practice must follow a completed trial. VoI methods are novel in that they explicitly direct research to support decision-making. Recent developments have facilitated VoI analyses, but my proposed short-term objectives will allow practitioners to use these methods while also making them more applicable to early stage trial designs and adaptive trials. My long-term goal of improving the use of VoI in research design through training, guidance and software will allow innovative trial methods that generate evidence to support decision- and policy-making. Within ConVOI, I will ensure this research is supported and disseminated through short-courses, presentations and tutorials.