Pharma R&D yield (number of approved drugs per year) is flat or trending slightly downward, and the cost per approved drug is skyrocketing. Even with powerful new pharmacometrics and statistical techniques and a keen focus on reducing late-stage attrition, the number of compounds in late stage development fell significantly in 2011. R&D productivity is not sufficient to support current spend rates.
The bar for getting drugs approved is higher than ever. Approvals now require direct proof of outcome. New drugs compete with existing, proven and known off-patent (and low-cost) drugs. New techniques generate a tsunami of data, but these have minimal impact, because their implications for human outcomes are not at all clear. Drug development is not just more difficult than it was ten years ago, its different.
No amount of good clinical practice will change flawed compounds we are testing in trials into safe, effective treatments. But even a slight improvement in the quality of our promotion decisions yields a big increase in R&D return. The critical decision that adds value in Pharma R&D is deciding which compounds go into the clinic.
The difficulty? We must choose compounds that are safe and effective in humans before we have direct human data. Statistically powerful descriptive techniques can be used on a subset of preclinical data, and adjusted to predict human outcomes. But preclinical and human physiology can be very different, and scaling and extrapolation adds uncertainty. Purely statistical approach are also unsuitable for integrating the petabytes of available data about a disease and related targets and compounds. "Mental models" attempt to account for all these data, but are not quantitative, and have questionable predictive power that appears to be declining in the face of increasingly complex diseases. We need scientific methods that can use more relevant data for better outcome predictions to support better translational decisions.
At Clermont, Bosley LLC we apply our proven PhysioMech™ modeling approach to collaborate with clients, creating quantitative models that can scientifically integrate large amounts of data and yield models that predict human trial (or animal experiment) outcomes prior to the trial, in a manner that is open, documented, provides insight and promotes common understanding and consensus. Once trial data are available, these model incorporate them to (for example) allow extrapolation to other patient demographics, identify biomarkers to make decisions faster and with less risk, or to optimize formulations. Clermont, Bosley LLC can help.
_
The bar for getting drugs approved is higher than ever. Approvals now require direct proof of outcome. New drugs compete with existing, proven and known off-patent (and low-cost) drugs. New techniques generate a tsunami of data, but these have minimal impact, because their implications for human outcomes are not at all clear. Drug development is not just more difficult than it was ten years ago, its different.
No amount of good clinical practice will change flawed compounds we are testing in trials into safe, effective treatments. But even a slight improvement in the quality of our promotion decisions yields a big increase in R&D return. The critical decision that adds value in Pharma R&D is deciding which compounds go into the clinic.
The difficulty? We must choose compounds that are safe and effective in humans before we have direct human data. Statistically powerful descriptive techniques can be used on a subset of preclinical data, and adjusted to predict human outcomes. But preclinical and human physiology can be very different, and scaling and extrapolation adds uncertainty. Purely statistical approach are also unsuitable for integrating the petabytes of available data about a disease and related targets and compounds. "Mental models" attempt to account for all these data, but are not quantitative, and have questionable predictive power that appears to be declining in the face of increasingly complex diseases. We need scientific methods that can use more relevant data for better outcome predictions to support better translational decisions.
At Clermont, Bosley LLC we apply our proven PhysioMech™ modeling approach to collaborate with clients, creating quantitative models that can scientifically integrate large amounts of data and yield models that predict human trial (or animal experiment) outcomes prior to the trial, in a manner that is open, documented, provides insight and promotes common understanding and consensus. Once trial data are available, these model incorporate them to (for example) allow extrapolation to other patient demographics, identify biomarkers to make decisions faster and with less risk, or to optimize formulations. Clermont, Bosley LLC can help.
_