Reusable Patterns: The Solution to Repeatable Endpoint Processes
I have observed many of the different adjudication workflows used across a variety of sponsor and CRO trials, and noticed that divergence in workflows is greatly influenced by issues of control and trust.
For example, a relatively simple two-adjudicator workflow followed by a CEC evaluation in cases of discordance can be made more complex when the CRO requires the completed voting eCRFs to be reviewed in-house before they are made available to the clinical event committee.
Whether this extra set of steps is necessary or not is not in question. What is true is that the addition of this set of steps generates data (e.g. when were the forms sent, who reviewed them, what was the outcome of the review, when was the dossier made available to the committee), and this data gets clumped together with the data generated by the disposition of the case under its normal workflow. This additional workflow snippet does not modify the two-adjudicator + CEC pattern: it is just a control overlay on top of it. Nonetheless, if it’s not recognized as such, it only serves to obfuscate the information that’s central to the premise of the trial in question.
I believe the answer to the validation + flexibility conundrum lies in reusability and patterns. Reusable patterns have been used in the field of software engineering for many years, and have injected a level of reliability and predictability when testing complex system implementations. Software Design Patterns can be augmented by Decorator Patterns. These decorator patterns are applied to enhance or add functionality to basic design patterns without necessarily changing their core behaviors.
I suggest that Event Patterns be defined to handle specific event types. These may include the two-adjudicator + CEC pattern, and many others such as an odd number of adjudicators with majority rule.
Additionally, Control Patterns should be defined as overlays on top of event patterns to add various kinds of audit/control related actions. System and user action logs, which record every interaction with and by the adjudication system, can then be reported either in the aggregate (as it’s done today), or by segregating actions depending on whether they were part of event management or control workflows. By enabling event and control patterns, sponsors and CROs retain the level of flexibility they need to carry out their studies.
Concurrently, being able to separate system interactions by pattern type enables sponsors and regulators to look at significantly clearer data, thus seeing structure where they previously saw noise.
Creating event and control patterns cannot be done reliably using traditional computer languages. The complexity is too high when expressing these concepts in Java, C++ or similar development environments. An electronic endpoint/event adjudication system (EAS) must be able to represent these patterns using higher level semantics (e.g. graphically) with facilities to place the control overlays and enable the decoupling of audit logs when necessary.
For the industry to be able to adopt this adjudication management functionality, the patterns must be defined and validated, and it should be clear that they are being used and implemented in the EAS without modification. This will reduce the time it takes to launch complex adjudication projects, will reduce the level of errors, and minimize the time it takes sponsors and regulators to determine the direction of experimental therapies.
Industries pursuing transformation need vision and aspiration. Data quality in clinical trials is paramount. Whether collecting patient scans as surrogate endpoints, or collecting patient data for serious event or endpoint adjudication, ensuring this data is devoid of errors can mean the difference between success and failure. We must remain focused on continued innovation and entrepreneurship in our industry. Delays and mistakes must be scrubbed out of the process. If we can agree on those principles, we can move even closer to zero-delay clinical trials.