diff --git a/Concept-Paper-2021-09-03.md b/Concept-Paper-2021-09-03.md index ca4e464..363d2fa 100644 --- a/Concept-Paper-2021-09-03.md +++ b/Concept-Paper-2021-09-03.md @@ -32,31 +32,35 @@ The objective of this research project is to estimate ## Potential Topics Some areas of interest. - 1. Estimate Probability of Success of moving between Phases. - 1. Depends on either FDA data or use of an ML matching process (Hard to do right). - 2. Follow E.K.’s approach of splitting Technical vs Scientific failures to progress. - 3. Denormalize the “process” of moving between phase 1 → phase 2 → phase 3 that is talked about and actually represent the real paths taken. - 1. As selection from a set of paths. - 2. Seeking approval for a new indication may not require initial phase trials. This “extraneous” information may be hard to capture/model/control for. - 3. Denormalizing this path may work as a proxy for changes in regulatory strictness? - 1. Probably not, because choice of path is probably dependent on already known information on safety. - 2. Estimate probability of success for approval based on trial data? How do biologics differ from standard small-molecule drugs? - 1. Haven’t put much thought in to this, not sure what the value proposition would be. - 3. Estimate entry rates/Probability Of Success of Biologics/biosimilars - 1. There appear to be lots of competing biosimilars in insulin. - 4. Develop “diagnostic tests” that will allow effects of policy to be forecast partway through implementation. - 1. Something such as measures of phase 1 trials and phase 2 trials to predict phase 3 & approval rates. -Possible Hypotheses +1. Estimate Probability of Success of moving between Phases. + 1. Depends on either FDA data or use of an ML matching process (Hard to do right). +2. Follow E.K.’s approach of splitting Technical vs Scientific failures to progress. +3. Denormalize the “process” of moving between phase 1 → phase 2 → phase 3 that is talked about and actually represent the real paths taken. + 1. As selection from a set of paths. + 2. Seeking approval for a new indication may not require initial phase trials. This “extraneous” information may be hard to capture/model/control for. + 3. Denormalizing this path may work as a proxy for changes in regulatory strictness? + 1. Probably not, because choice of path is probably dependent on already known information on safety. +2. Estimate probability of success for approval based on trial data? How do biologics differ from standard small-molecule drugs? +1. Haven’t put much thought in to this, not sure what the value proposition would be. +3. Estimate entry rates/Probability Of Success of Biologics/biosimilars +1. There appear to be lots of competing biosimilars in insulin. +4. Develop “diagnostic tests” that will allow effects of policy to be forecast partway through implementation. +1. Something such as measures of phase 1 trials and phase 2 trials to predict phase 3 & approval rates. + + +## Possible Hypotheses Do Probabilities of Success differ significantly when estimated on normalized or denormalized paths? - • With the FDA data, this will be uniquely answerable. Not sure though what those implications of interest are though. - • The value proposition here is that it might “measure” the risk reducing effect of that extra knowledge from previous studies, evidence, etc. I wonder if this could act as an instrument on the effect of unknown risk when starting a new compound? - ◦ The underlying assumption is that the path depends heavily on whether the compound has been studied previously or not. That may not be true. - ◦ Provides the “risk value” of prior phases. Maybe allows modeling as net future value? + +• With the FDA data, this will be uniquely answerable. Not sure though what those implications of interest are though. +• The value proposition here is that it might “measure” the risk reducing effect of that extra knowledge from previous studies, evidence, etc. I wonder if this could act as an instrument on the effect of unknown risk when starting a new compound? + ◦ The underlying assumption is that the path depends heavily on whether the compound has been studied previously or not. That may not be true. + ◦ Provides the “risk value” of prior phases. Maybe allows modeling as net future value? Do Probabilities of Success depend on reimbursment rate? - • The thought here is that if reimbursement rates increase, more R&D is spent on “long-shot” drugs that are likely to develop new fields of research. - • Requires some sort of reimbursement/pricing data. Not necessarily available, and would require a natural experiment. +• The thought here is that if reimbursement rates increase, more R&D is spent on “long-shot” drugs that are likely to develop new fields of research. +• Requires some sort of reimbursement/pricing data. Not necessarily available, and would require a natural experiment. Intermediate goals + 1. Find literature that describes each difficult section of the R&D to successful launch process, focusing on the second section. Policy questions/ topics @@ -65,14 +69,15 @@ Precision medicines & trials Insurers paying for participants in a clinical trial. Medicare pays for many, but what effect does this have on recruitment? Possible variation across drugs (legislatively), as identification. - • Endpoint doesn’t matter for recruiting - • Our data may show the “rate” at which people join trials. - ◦ Does medicare coverage for the trial speed that up? +- Endpoint doesn’t matter for recruiting +- Our data may show the “rate” at which people join trials. + - Does medicare coverage for the trial speed that up? Use of surragate endpoints. Risk sharing between insurer and developer. - • Cancer: survival rates vs what happened to the cancer - • Reduces length of trial, but requires phase 4 trials - • Effects on recruiting/timing + +- Cancer: survival rates vs what happened to the cancer +- Reduces length of trial, but requires phase 4 trials +- Effects on recruiting/timing Possible Tasks Lit review