Proposal | Current state | Barriers to implementation | Potential solutions |
---|---|---|---|
Think Bayesian | Nearly all clinical trials are designed based on frequentist statistics, which yields less information about the probability of benefit and harm than Bayesian statistics | Unfamiliarity with the Bayesian framework and adaptive trial design Skepticism about the perceived subjectivity of Bayesian prior distributions | Education for clinicians, investigators, and funders Widespread appreciation for the close analogy between Bayesian statistics and routine clinical reasoning |
Adapt when needed | Standard clinical trials employ rigid designs with fixed sample sizes based on educated estimates of event rates and treatment effect | Skepticism about the statistical validity of frequent interim analyses to guide adaptations in design Methodological expertise in adaptive trial design and Bayesian trial design is not yet widespread | Dissemination of consensus on best practices for the design and conduct of adaptive trials Successful completion and publication of arms and domains from ongoing adaptive trials |
Build a platform | Traditional clinical trials typically build trial infrastructure (funding, regulatory approval, logistics, recruitment, investigator network, analysis, data monitoring committee) to address one research question | Funding is complex: Both the platform infrastructure and the individual interventional research questions require funding support Authorship criteria can be challenging to establish | Proactive partnerships with funding agencies to support platform infrastructure Academic community should intentionally place greater value on group authorship |
Understand the noise | Some trials are criticized for being too pragmatic and providing fixed interventions regardless of mechanistically relevant physiological or biological characteristics of individual patients | Demonstrating heterogeneity of treatment effect increases sample size requirements Clinical importance of different mechanistic pathways may not be clear prior to the conduct of the trial | Development and validation of short-term surrogate endpoints reflecting mechanistically relevant treatment response for phase II trials (similar to phase II oncology trials) before evaluation in phase III trials Use of response-adaptive randomization and other adaptive trial techniques to enhance recognition of differential treatment response among relevant patient subgroups |
Be inclusive | Most patients in critical care units are not enrolled in a clinical trial Many trials do not include patients from resource-constrained settings | Restrictive inclusion and exclusion criteria Challenges of obtaining informed consent in timely fashion for time-dependent interventions | Create potential incentives to including patients in clinical trials Consider enrolment in trials as a quality performance marker for healthcare systems Enhance patient engagement in trial design |
Embed discovery within care | Trials employ dedicated electronic case report forms and data coordinating centers Data entry is often duplicated across multiple interfaces, increasing both workload and errors | Widely accepted dissociation between clinical care and clinical research. Widespread variability in electronic health record systems Resource-constrained settings may not have access to electronic health record systems Algorithms for detecting and reducing missing data are not routinely implemented outside of trials | Partner with healthcare administrators to design trials that contribute to quality and innovation within healthcare systems Integrate clinical record-keeping with minimum datasets for clinical research Develop low-cost electronic health record systems for widespread use |