According to Gen Li, president of Phesi, limited thinking and outdated trial models leave studies bound for failure. This approach, he said, contributes to the approximately 33% failure rate of studies, as well as about one-third needing significant protocol amendments.
The company recently tackled a comprehensive analysis of 330,000 clinical trials from the past 15 years. According to this analysis, most employed a primitive approach that unnecessarily lengthened the trial process, incurred additional costs, and increased the risk of failure.
Outsourcing-Pharma recently spoke to Li about the risks in using outdated study designs, how a modal-value approach could avoid many issues, and how to rethink design.
OSP: Could you please share some perspective on what makes a clinical trial “primitive” versus modern?
GL: A primitive approach effectively means relying on gut instinct and emotion, rather than data. So a sponsor might make decisions based on their own personal history to lead the design of a protocol or those of other opinion leaders. A modern approach takes the guesswork out of the equation and uses integrated data science approaches, including AI/ML/NLP, to provide insights into trial design that are definitively supported by data; this allows the thought leaders and scientists to work with real evidence to help support their collective goals.
OSP: Could you please share some reasons why trial teams might still be leaning on primitive models?
GL: Life sciences organizations and trial sponsors are notoriously resistant to change. The indisputable value to be derived from data is not lost on sponsors, and the pandemic has been a catalyst, but change happens incrementally.
The change process here forces sponsor companies into a change management environment that crosses functional silos. And it is a highly educational process that shows the difference between what is being done, and what can be done.
The instinct of many sponsors to try to better manage data and technology on their own, but they find the first and biggest hurdle to be data itself. Not only accessing the right data, but de-siloing their own data lakes, harmonizing with external and unstructured sources, and applying algorithms and models that are both domain-specific and prove consistently and systematically. They often invest heavily in getting the data and building analytics to support their clinical development. But still within the traditional framework of approaching protocol design, rather than being guided by the data – as is the case with modal values-led design.
Finally, another reason for resistance to adoption is that, in reality, it can be hard for organizations to accept that a third party would have greater data volume, variety, and velocity than they have internally. It is uncomfortable for R&D leaders to admit that another expert may know more about your project or set-up than you do yourself.
Organizations have built huge systems, processes, and functions around protecting their data, only now to find that small, tech-savvy firms and start-ups can do more with their data than they can themselves, driven by the rapidly evolving power of AI and big data technologies in extracting and structuring data. A change in mindset is needed if sponsors are to accelerate trials and most importantly, get treatments to patients more quickly.
OSP: Can a primitive trial that already is underway be adjusted?
GL: Yes, rescuing a trial is perfectly possible. Data science techniques can be applied mid-stream for course correction of a trial, which avoids repeated protocol amendments.
Avoiding costly protocol amendments that add length to a trial is one of the main reasons why leveraging data early in the clinical development lifecycle is so important. The commonality of protocol amendments points to a need for the data science-led approach.
OSP: How does a primitive design approach create the challenges you mentioned (increased costs, longer trials, enrollment challenges, etc)?
GL: Without an integrated approach to predictive science to support a trial, enrollment will take longer and be less successful in the first instance because of a lack of available patients. Slow enrollment adds length to a trial which greatly increases costs. Protocol amendments are also a result of primitive design – the estimated cost of which to a trial is around $500,000.
Ultimately, not taking a data-led approach results in failed trials. Everyone loses when a trial fails – research subjects, regulators, sponsors, and critically, patients.
OSP: Could you please share some key pieces of advice for making the switch to the more effective, efficient model-value approach?
GL: What’s needed is an integrated approach to predictive science across the clinical development plan, protocol design optimization, synthetic control arms, and investigator selection.
The first step is to make sure there is structured data available for your specific areas of interest. Secondly, ensuring that there is an analytic framework around that data so that you can interrogate it in different ways to suit the specific questions around your study design and patient population.
OSP: Do you have anything to add?
GL: A modern approach from the beginning of clinical development can help assure patient centricity. It has a direct impact on the commercial end of the sponsor, as well as the patients and research subjects along the way. The upside from adopting a more modern outlook will be profound to both patients and the industry.
Trials should be far less expensive and time and resource-consuming, with less burden on the research subjects and the research sites. This is achievable when sponsors take a data-led approach to trials.