As it now costs a whopping $800 million to develop a new drug, American researchers have developed a forecasting computer model which could reduce drug development costs, saving hundreds of millions of dollars per new drug. Their Bayesian network model is based on publicly available data about 500 successful and failed new drugs. And they say that the application of their model would reduce mean capitalized expenditures by an average of $283 million per successful new drug (from $727 to $444 million). Now it remains to be seen if the pharmaceutical industry will use this forecasting tool.
This Bayesian model has been built by Asher Schachter and Marco Ramoni, researchers at the Children's Hospital Informatics Program (CHIP), a multidisciplinary applied research program at Children's Hospital Boston and the Harvard-MIT Division of Health, Sciences and Technology.
[They] constructed a Bayesian network model to calculate the probability that a given new drug would pass successfully through Phase III trials and receive New Drug Application (NDA) approval. Their approach differs from convention in modeling populations of drugs rather than populations of patients. They used publicly available safety and efficacy data for about 500 successful and failed new drugs, broken down by therapeutic category, then confirmed the validity of their model by testing it with a group of cancer drugs whose fates are already known.
Below is a chart showing the pharmaceutical industry performance for delivering new drugs when analyzed with this computer model (Credit: Asher Schachter and Marco Ramoni). Here is a link to a larger version.
The model was validated on an independent data-set consisting of successful and failed drugs for one class (antineoplastics), and found to perform with 78% accuracy (80% sensitivity and 76% specificity)12. Entering hypothetical models with a range of sensitivity and specificity values into the same Monte Carlo simulation framework used to assess the Bayesian network model showed that model sensitivity and specificity values of only 60% or better demonstrate a potential financial benefit over the recent reported performance of the pharmaceutical industry.
And what are the results produced by this model?
This analysis, using summary data on industry-reported expenditures and revenues, indicated that application of the model would reduce mean capitalized expenditures by an average of $283 million per successful new drug (from $727 to $444), and increase revenues by an average of $160 million per Phase III trial (from $347 to $507 million) during the drug's first seven years on the market.
The researchers also claim that the application of their model would be beneficial to everyone and not only to the pharmaceutical industry.
[They think] that more accurate clinical forecasting would eliminate unsafe investigational new drugs; avoid subjecting patients to unnecessary drug trials; reduce the cost of prescription drugs for consumers; and empower the industry to take risks on truly innovative new drugs, so that more get to market.
This research work has been published by Nature Reviews Drug Discovery under the name "Clinical forecasting in drug development" (Volume 6, Issue 2, Pages 107-108, February 2007). Here is a link to the article. The above illustration comes from a sidebar article about their application of a Bayesian model for clinical forecasting.
Finally, it's interesting that the two researchers behind this project are also businessmen and founders of Phorecaster, a company "focused on predictive modeling in drug development." So this research work is also a promotional tool for their commercial activities.
Sources: Children's Hospital Boston news release, February 1, 2007; and various other websites
You'll find related stories by following the links below.