Probability of success (POS) of a clinical trial is a critical metric for clinical researchers and biopharma investors when making scientific and economic decisions. Unfortunately, many of the conventional tools used to estimate approval rates and predict approvals are typically based on a small number of examples that have incomplete information and do not acknowledge the valuable insights that can be gleaned from partially reported investigations.
A research team at the MIT Laboratory of Financial Engineering, using datasets provided by Informa Pharma Intelligence, has applied machine learning techniques to train and validate its predictive models taking into account predictive factors such as drug compound characteristics, clinical trial design, previous trial outcomes and the sponsor track record.
“More accurate risk metrics will eventually lead to fewer big failures, faster approval times, cost savings to the entire healthcare system and more investment capital for developing breakthrough therapies. We believe our approach is more accurate and more relevant from an investment perspective than previous ones because we are using a larger dataset than anyone has ever used,” MIT’s Andrew Lo told Scrip.
More importantly, the analytics tool -- dubbed Project Alpha (Analytics for Life-sciences Professionals and Healthcare Advocates) -- is not as static as other approaches and is continuously updating. Indeed, a recent uptick in the likelihood of better clinical outcomes was identified. Compared with previous POS estimates, Lo and his colleagues obtained higher POS estimates for all phases of development. Use of biomarkers in recent years, especially for patient stratification, has contributed to improved success rates.
MIT and Informa are making the clinical success rate metrics available to industry stakeholders via the LFE-hosted website http://projectalpha.mit.edu, which will provide regularly updated aggregate clinical success rates and durations, as well as disaggregated estimates across trial features such as disease type, clinical phase, time and lead indication status. Potential collaborators will also be able to explore opportunities to enhance the datasets that underpin the analytics.
“We hope and expect to engage with all stakeholders from the biomedical ecosystem: investors, biopharma executives, regulators, clinicians, scientists, patients and their advocates. Each of these stakeholders can provide critical value to our overall goal of developing more accurate analytics to support biomedical innovation,” he added.
To encourage additional analyses using the metrics and data underpinning them, Informa is making its commercial datasets available to academic researchers through an academic licensing program and a research proposal review process.
Lo believes the most immediate beneficiaries of the new tools will be biotech investors and drug developers as accurate evaluation of a candidate drug’s likelihood of approval is critical to the efficient allocation of capital. “Eventually, we expect these tools to benefit regulators, scientists, and ultimately desperate patients waiting for life-saving therapies that drug developers have the expertise to create but lack the resources and/or financial motivation to do so,” he added.