Collecting competitive data isn’t the problem; most biopharma companies are drowning in it. But making sense of the data, using it to predict likely shifts in the competitive landscape and drive decisions can be a challenge.
While forecasting drug timelines and success rates are critical to inform investment and R&D decisions, the standard approach has limitations that could expose companies to more risk and reduce their ROI.
But, increasingly, biopharma companies, looking to advance from a reactive stance to a predictive approach – see the progression mapped out on the Cortellis Competitive Intelligence Maturity Curve – are turning to a new toolbox that includes machine learning, mathematical modeling and predictive analytics. This allows a company to sharpen its strategy, set priorities and make smarter, quicker decisions about its portfolio.
Implementing new approaches
The competitive information is out there – on thousands of companies, tens of thousands of assets in development, progress in clinical trials, financial and corporate history. But how can a biopharma company capitalize on that vast source of information effectively? How can it draw insights from the data? Is there a methodology for utilizing that information for predictive modeling and analytics?
Integration of data. A first step in the challenge is the integration of heterogeneous data to create insights; bringing the data together in one place makes it far more valuable.
Visualizing novel industry trends and predicting shifts in the strategic landscape. Bringing the data together in the right cohesive format is a powerful step, but just having the raw data is not going to be sufficient. Can it be used to visualize coming trends, or reliably predict what’s going to happen in the future? Can it pinpoint the shifts in time? Can it envision the coming breakthroughs? Can it foresee changes in a complex competitive landscape?
The importance of ‘failing early’
Identifying strategic opportunities early. This is critical to managing a portfolio properly. Not only can a company prioritize the most promising drug development opportunities at an early stage, it can also make early decisions to discontinue programs that seem heading toward a dead-end. “Failing early,” in turn, allows the company to strategically shift finite R&D resources to more promising opportunities in its portfolio.
Companies can transform how they forecast timelines and drug success rates when they use a predictive analytics approach.”
The Cortellis predictive model, Cortellis Analytics – Drug Timeline & Success Rates, utilizes 15 qualitative traits that are known to correlate with the successful progression of most drug development programs. The model has been built so that additional traits can be added in the future and integrated into the model based on experience as well as client needs. Further, the model can mature based on competitive landscape changes and on the evolution of the industry. For example, the dramatic impact of recent regulatory designations aimed at accelerating development programs to early approval would not have been included in the model had it been built just five years ago.
No two drugs carry these traits with the same weight. Each of the 15 qualitative traits has different weights and contributes differently to the predictive model based on the stage of the drug’s development. The Cortellis predictive model doesn’t just look at benchmarks and qualitative traits, but also adjusts itself to future milestones based on their outcome for each of the drug programs.
Companies can transform how they forecast timelines and drug success rates when they use a predictive analytics approach:
- Machine learning is applied to identify paths and time to approval for drug development programs.
- Predictions are generated for each drug development project, from phase I and above, by considering the projected impact of each aspect of that project.
- An expansive set of historical and current data can be pulledinto the model.
- Results are based on quantifiable, data-driven evidence derived from extensive data preparation and statistical modelling.
- Outcomes are adjusted to critical milestones, so predictions are based on current information rather than a snapshot in time.
The Cortellis predictive analytics model was introduced in September at the Pharma CI USA Conference & Exhibition in Newark, N.J., at a session titled “Make Better Pharmaceutical Portfolio Decisions – Using Failure Rates to Understand Portfolio Risks and Improve R&D Productivity.”
Join the author, Karthik Subramanian, as well as Jamie Munro, Global Head, Portfolio & Licensing at Clarivate Analytics, for an encore presentation of the talk via webinar on Oct. 18. For more information and to register, click here.
