> Although an adaptive design guidance finalized in 2019 left the door open for bayesian trials, their use in drug development has to date been limited,4 such as in Ebola and SARS-CoV-2 epidemics and in pediatric and rare disease trials.
the reason it has been limited to those cases is
drug development, today, is constrained by commercialization.
all four categories listed - ebola, sars-cov-2, pediatric and rare disease drugs - each for their own reason, have low commercialization risk, so if they are scientifically robust solutions, there is ROI.
most drugs in development today are new indications for existing molecules, generally in oncology, also because these are the most favorable conditions for commercialization.
commercialization, commercialization, commercialization. people don't want innovations in drug approval, nobody is clamoring for that. they want innovations in commercialization.
i don't know if bayesian methods will make trials a lot cheaper. and that's their problem! there are already a lot of very smart people working on this issue, and they have litigated to death all the objective facts.
There's a weird thing that happens with cancer drugs that I just experienced.
When they are working with a pharmacy and your insurance, the price they'll charge you is $10,000 for enough pills to last a month. But when your insurance says "We won't cover that" all the sudden you find out the company has a backdoor subsidization program which will fully cover the cost of the drug for reasons I can't really fathom (good will?)
What's even more uncomfortable is my insurance (aetna) also mandates that I get my medicine from their subsidiary, CVS.
I really don't like this sort of thing. The price of everything in medicine feels distorted in unbelievable ways. Like famously a $0.25 acetaminophen pill that somehow magically costs $10 in the hospital. I guess there's some nice individual packaging.
I'm less and less convinced that it's even a cost sort of thing with these pharmaceutical companies. Like, sure they'll love to reduce that as much as possible. But the price itself seems entirely fictitious and based on what they can commonly get insurers to pay, and not anything related to the actual R&D of the drug.
> the reason it has been limited to those cases is drug development, today, is constrained by commercialization.
That's a good observation, but I think it's an incomplete picture. Another important constraint is often regulatory inertia and historical baggage.
The UK pioneered small classical and adaptive trials using Bayesian methods, and there were some promising results. A lot of modern Bayesian methodology was, in fact, developed at the MRC BSU Cambridge with this goal in mind. For example, the probabilistic programming language BUGS (1989).
Given that most drugs fail, the industry is highly incentivized to use Bayesian methods to fail faster. These models allow for more rapid dose-finding and the ability to distinguish promising leads using interim data, which is vital given the massive cost of any trial, especially late-stage failures.
But for Bayesian methods to make a dent, they'd need to be applied to a large number of trials, and change doesn't happen overnight. Lots of big pharma players, e.g. GSK, are becoming interested in moving to Bayesian methods in order to leverage prior information and work better within small-data regimes.
> Bayesian inference assumes the observed data are fixed and aims to quantify the evidentiary support for all possible levels of treatment effectiveness based on the data at hand.
The problem with this approach is that we can only observe ONE level of treatment effectiveness, i.e., the level of treatment effectiveness that the treatment actually possesses. All other possible levels of effectiveness are entirely hypothetical. There's no data about all these other possible levels of effectiveness because they don't occur in reality. So the data cannot possibly tell you anything about how likely is the observed outcome, because the observed outcome is the only outcome that you observe. I
This criticism was made over 100 years ago, and Bayesians still don't have an answer. They just keep going as if nothing happened, but the reality is their methodology is utterly and fatally flawed.
18 comments
> Although an adaptive design guidance finalized in 2019 left the door open for bayesian trials, their use in drug development has to date been limited,4 such as in Ebola and SARS-CoV-2 epidemics and in pediatric and rare disease trials.
the reason it has been limited to those cases is
drug development, today, is constrained by commercialization.
all four categories listed - ebola, sars-cov-2, pediatric and rare disease drugs - each for their own reason, have low commercialization risk, so if they are scientifically robust solutions, there is ROI.
most drugs in development today are new indications for existing molecules, generally in oncology, also because these are the most favorable conditions for commercialization.
commercialization, commercialization, commercialization. people don't want innovations in drug approval, nobody is clamoring for that. they want innovations in commercialization.
i don't know if bayesian methods will make trials a lot cheaper. and that's their problem! there are already a lot of very smart people working on this issue, and they have litigated to death all the objective facts.
When they are working with a pharmacy and your insurance, the price they'll charge you is $10,000 for enough pills to last a month. But when your insurance says "We won't cover that" all the sudden you find out the company has a backdoor subsidization program which will fully cover the cost of the drug for reasons I can't really fathom (good will?)
What's even more uncomfortable is my insurance (aetna) also mandates that I get my medicine from their subsidiary, CVS.
I really don't like this sort of thing. The price of everything in medicine feels distorted in unbelievable ways. Like famously a $0.25 acetaminophen pill that somehow magically costs $10 in the hospital. I guess there's some nice individual packaging.
I'm less and less convinced that it's even a cost sort of thing with these pharmaceutical companies. Like, sure they'll love to reduce that as much as possible. But the price itself seems entirely fictitious and based on what they can commonly get insurers to pay, and not anything related to the actual R&D of the drug.
> the reason it has been limited to those cases is drug development, today, is constrained by commercialization.
That's a good observation, but I think it's an incomplete picture. Another important constraint is often regulatory inertia and historical baggage.
The UK pioneered small classical and adaptive trials using Bayesian methods, and there were some promising results. A lot of modern Bayesian methodology was, in fact, developed at the MRC BSU Cambridge with this goal in mind. For example, the probabilistic programming language BUGS (1989).
Given that most drugs fail, the industry is highly incentivized to use Bayesian methods to fail faster. These models allow for more rapid dose-finding and the ability to distinguish promising leads using interim data, which is vital given the massive cost of any trial, especially late-stage failures.
But for Bayesian methods to make a dent, they'd need to be applied to a large number of trials, and change doesn't happen overnight. Lots of big pharma players, e.g. GSK, are becoming interested in moving to Bayesian methods in order to leverage prior information and work better within small-data regimes.
> Bayesian inference assumes the observed data are fixed and aims to quantify the evidentiary support for all possible levels of treatment effectiveness based on the data at hand.
The problem with this approach is that we can only observe ONE level of treatment effectiveness, i.e., the level of treatment effectiveness that the treatment actually possesses. All other possible levels of effectiveness are entirely hypothetical. There's no data about all these other possible levels of effectiveness because they don't occur in reality. So the data cannot possibly tell you anything about how likely is the observed outcome, because the observed outcome is the only outcome that you observe. I
This criticism was made over 100 years ago, and Bayesians still don't have an answer. They just keep going as if nothing happened, but the reality is their methodology is utterly and fatally flawed.