Failed clinical trials can cost sponsors more than a billion dollars, and waste years of time developing a drug that will never get to market. However, these losses can now be mitigated through the use of artificial intelligence (AI) and machine learning-driven platforms that identify sub-populations of patients within a clinical trial who could respond positively to a treatment.
When sponsors can find these populations and adapt the trial design accordingly, it can reduce the rate of trial failure, rescue a promising molecule for further development, and capture better primary and secondary endpoint data to support approval and payer valuations.
This paper explores how sponsors can leverage AI driven tools, such as IQVIA’s Sub-Population Optimization Solution (SOS), throughout the trial lifecycle to reduce the risk of failure, while improving success rates, and identifying new opportunities in drug development.
PART 1: THE HIGH COST OF FAILURE
The enormous cost of drug development is caused in large part by the high rate of failure. A 2019 study found that less than 14 percent of all drug development programs eventually lead to approval, and that number drops to just six percent for orphan drugs, and a mere 3.4 percent for oncology drugs, which is a dominant category for drug development.
When trials get shut down, sponsors face enormous financial losses — an estimated $800 million to $1.4 billionii — and patients lose the hope of potentially lifesaving treatments coming to market.
While some molecules simply don’t pan out, many trials fail because the trial design did not adequately identify the appropriate population, endpoint, and/or dose selection. This can result in a majority of trial participants not responding as hoped, or adverse events occurring that make safety risks too high. However, with the right data, this high rate of failure doesn’t have to occur.
Within failed trials, there are often sub-populations of patients who respond well to the treatment and suffer no adverse events, but because the overall numbers don’t add up, the trial is abandoned. This means sponsors lose their investment and patients who could potentially benefit from lifesaving treatments are denied these innovations.
These populations can be identified through the use of SOS platforms, which leverage AI and machine learning tools that can rapidly analyze clinical data sets to identify the most promising populations, allowing sponsors to adapt the trial accordingly. When sponsors can identify these sub-populations sooner in the research process, they can adapt the trial to focus on those patients. This increases safety and treatment outcomes through better trial design and can rescue promising trials that would otherwise be deemed failures.
PART 2: HOW THE TECHNOLOGY WORKS
These statistically rigorous analytical platforms feature machine learning algorithms that are programmed to analyze trial data to identify predictive biomarker(s) in treatment populations who are experiencing stronger outcomes. They can be used at various points in the trial lifecycle to identify trends and outliers without bias, to inform trial design, support approval and pricing decisions, and shape market and sales strategies.
The platform utilizes an industry recognized analysis method called Subgroup Identification based on Differential Effect Search (SIDES). SIDES is used to evaluate trial results to help clients answer tough questions about the effectiveness of a drug in development, and to develop effective trial design strategies to maximize their chances of success. These platforms use secure data management technologies and intuitive workflows that automatically query patient data sets based on predetermined features, and deliver results in easy-to-read charts, scatter plots, and summaries.
The easy to read graphical interface allows stakeholders to quickly review and understand which patient populations are responding to the treatment, who among them are facing a higher risk of adverse events, which populations are not likely to benefit, and where there are cut offs. For example, an analysis might show in a series of graphics that patients over the age of 65 with stage three cancer are more likely to respond to the treatment than younger patients, and that those in earlier stages of the cancer have a higher rate of adverse events.
This kind of instant easy to understand data gives investigators and sponsors the insights they need to adapt the trial and select the right patients with greater precision and confidence. That can reduce time, cost and risk across the trial lifecycle, from early discovery through commercialization, and enable sponsors to bring more successful treatments to market.
And when the data shows that no population is having meaningful outcomes, this technology can also save sponsors hundreds of millions of dollars by shutting down failing trials sooner and moving those resources to molecules that have a greater potential of generating returns.
These platforms deliver value at many stages of the trial lifecycle from early stage decision making, through approval, market access and extension studies