3 ms·

>> controlled clinical trials (which are still the only way to determine if a new drug works) His objections are good, but this mindset will probably have to g

by Jach 8y ago

>> controlled clinical trials (which are still the only way to determine if a new drug works) His objections are good, but this mindset will probably have to go in order to make headway on more efficient medicine. It would also alleviate the concern in #3, because once you give up that mindset, you actually have tools besides the (expensive) controlled clinical trial that allow you to make statements like "dried grape juice pills won't do a thing for Alzheimer's".

eesmith 8y ago
Do you have any examples of alternative methods to determine if a new drug works, or is this simply an expression of optimism? The pharmaceutical industry would pay billions for a cheaper or more effective replacement.
- Jach 8y ago
  The main alternative method I had in mind was that of causal inference. I can't recall seeing a very detailed paper of its effectiveness at judging a drug from observational data rather than a controlled (and expensive) clinical study but I'm sure it's been done, it's used in a number of areas in medicine and the social sciences. But it's more sophisticated than the classical battery of statistical tests you get to use if you design a classical clinical trial. However it works with clinical trials too -- it would probably help out with the replication crisis... The other method I had in mind is to put more focus on basic theory / models. Models in the sense of individual elements that can take input and produce output, not aggregated statistical models. These models can inform (and be informed by) later data. They're pretty useful with and without controlled experiments, e.g. either particle physics or astronomy. When applied to drugs, a functional model of how a drug would interact with the organism will tell you whether a drug works or not -- clinical trials and observation both can help refine the model as its outputs match up against reality. Of course good models of biology are hard, many mechanisms are poorly understood, but you need more focus at this stage before you even consider getting to the clinical trial. Unfortunately less effort seems to be spent on the modeling, more effort on getting quick iterative feedback via petri dishes then animal studies, and then disappointment when something fails in expensive human trials. With a lot more focus on the modeling side, with more willingness to use the data from experiments that aren't "gold standard" controlled clinical trials and to use tools like causal analysis, is it unreasonable to imagine that a lot less money could be spent for a lot sounder results? I'm not sure we can conclude such broad things about what the pharmaceutical industry wants or would and would not do. It's a mess of a beast. With the side effect of astronomical costs that results from this mentality of thinking controlled clinical studies are always necessary and the only way to know for sure, and more directly that you need them for every new drug to get to market because of regulations, is it any wonder that most new drugs are "me-too" drugs that companies "researched" from existing drugs to be just different enough to earn a new patent and get past the clinical trials that its predecessors did so they can sell it and make a profit? (Rent-Seeking and Redundancy, http://levine.sscnet.ucla.edu/papers/imbookfinal09.pdf http://levine.sscnet.ucla.edu/papers/imbookfinal09.pdf)
  - eesmith 8y ago
    I believe you are describing epidemiological studies. Some of those studies are already done. Viagra's effect on erectile dysfunction was based on collecting data about side-effects when evaluating its use as a treatment for angina. But I don't see how they can be applied to most drug tests. For example, in the news recently we learned of a new treatment which lead to a "complete response in breast cancer patient." https://www.eurekalert.org/pub_releases/2018-06/nci-nat060118.php https://www.eurekalert.org/pub_releases/2018-06/nci-nat06011... This requires an intervention and special personalized analysis of DNA and RNA from both a tumor and her normal tissue. There won't have been any observational data which lead to Taxol because it's not something that cancer patients would normally inject into themselves, outside of a clinical study. Most new drugs require complex and expensive synthesis, many may require special treatment methods or schedules, and most are for uncommon diseases. I don't see how a observational data from a non-clinical study can be effective. Similarly, radiation therapy, chemotherapy, and other such treatments are not normally found outside of clinical studies, and I don't see how it could be otherwise. As for basic theory / models, I follow the blog "In the Pipeline", by Derek Lowe, a pharmaceutical chemist. Links to his blog come up not infrequently here on HN. I get the very strong impression that models are used all the time, that model building is an important part of pharmaceutical chemistry, and the reason for iterative testing in assays, cell cultures, and the like is because the mathematical models aren't that good, even after decades of attempts. A DDG search for 'Boldrin' on that blog shows several articles which reference "Against Intellectual Monopoly", including http://blogs.sciencemag.org/pipeline/archives/2006/04/02/down_with_patents_eh http://blogs.sciencemag.org/pipeline/archives/2006/04/02/dow... . I am too ill-informed to be able to provide any commentary on my own. As I understand it, most of the "me-too" drugs are because multiple drug companies are working with relatively few drug targets. If company X comes out with a drug, and company Y comes out with a 'me-too' drug two years later, most likely what happened was that companies X and Y started looking for a drug some 15-20 years ago against the same target, and found similar compounds. Then when X got its patent, Y had to figure out some way to extract profit from its decade of work, which would otherwise go down the drain. It's not so much that Y "researched" X's existing drug to make a me-too compound. In addition, just because some drug appears to be a "me-too" drug, that doesn't make it useless. People's bodies are different. Quoting from a comment at http://blogs.sciencemag.org/pipeline/archives/2004/08/31/me_too_part_two http://blogs.sciencemag.org/pipeline/archives/2004/08/31/me_... , "Some people may not respond to Zocor but may respond beautifully to Lipitor. Or they may experience less muscle weakness under Crestor." See also http://blogs.sciencemag.org/pipeline/archives/2004/02/29/more_from_the_metoo_front http://blogs.sciencemag.org/pipeline/archives/2004/02/29/mor... .

by Jach 8y ago

eesmith 8y ago

Do you have any examples of alternative methods to determine if a new drug works, or is this simply an expression of optimism? The pharmaceutical industry would pay billions for a cheaper or more effective replacement.

Jach 8y ago
The main alternative method I had in mind was that of causal inference. I can't recall seeing a very detailed paper of its effectiveness at judging a drug from observational data rather than a controlled (and expensive) clinical study but I'm sure it's been done, it's used in a number of areas in medicine and the social sciences. But it's more sophisticated than the classical battery of statistical tests you get to use if you design a classical clinical trial. However it works with clinical trials too -- it would probably help out with the replication crisis... The other method I had in mind is to put more focus on basic theory / models. Models in the sense of individual elements that can take input and produce output, not aggregated statistical models. These models can inform (and be informed by) later data. They're pretty useful with and without controlled experiments, e.g. either particle physics or astronomy. When applied to drugs, a functional model of how a drug would interact with the organism will tell you whether a drug works or not -- clinical trials and observation both can help refine the model as its outputs match up against reality. Of course good models of biology are hard, many mechanisms are poorly understood, but you need more focus at this stage before you even consider getting to the clinical trial. Unfortunately less effort seems to be spent on the modeling, more effort on getting quick iterative feedback via petri dishes then animal studies, and then disappointment when something fails in expensive human trials. With a lot more focus on the modeling side, with more willingness to use the data from experiments that aren't "gold standard" controlled clinical trials and to use tools like causal analysis, is it unreasonable to imagine that a lot less money could be spent for a lot sounder results? I'm not sure we can conclude such broad things about what the pharmaceutical industry wants or would and would not do. It's a mess of a beast. With the side effect of astronomical costs that results from this mentality of thinking controlled clinical studies are always necessary and the only way to know for sure, and more directly that you need them for every new drug to get to market because of regulations, is it any wonder that most new drugs are "me-too" drugs that companies "researched" from existing drugs to be just different enough to earn a new patent and get past the clinical trials that its predecessors did so they can sell it and make a profit? (Rent-Seeking and Redundancy, http://levine.sscnet.ucla.edu/papers/imbookfinal09.pdf http://levine.sscnet.ucla.edu/papers/imbookfinal09.pdf)
- eesmith 8y ago
  I believe you are describing epidemiological studies. Some of those studies are already done. Viagra's effect on erectile dysfunction was based on collecting data about side-effects when evaluating its use as a treatment for angina. But I don't see how they can be applied to most drug tests. For example, in the news recently we learned of a new treatment which lead to a "complete response in breast cancer patient." https://www.eurekalert.org/pub_releases/2018-06/nci-nat060118.php https://www.eurekalert.org/pub_releases/2018-06/nci-nat06011... This requires an intervention and special personalized analysis of DNA and RNA from both a tumor and her normal tissue. There won't have been any observational data which lead to Taxol because it's not something that cancer patients would normally inject into themselves, outside of a clinical study. Most new drugs require complex and expensive synthesis, many may require special treatment methods or schedules, and most are for uncommon diseases. I don't see how a observational data from a non-clinical study can be effective. Similarly, radiation therapy, chemotherapy, and other such treatments are not normally found outside of clinical studies, and I don't see how it could be otherwise. As for basic theory / models, I follow the blog "In the Pipeline", by Derek Lowe, a pharmaceutical chemist. Links to his blog come up not infrequently here on HN. I get the very strong impression that models are used all the time, that model building is an important part of pharmaceutical chemistry, and the reason for iterative testing in assays, cell cultures, and the like is because the mathematical models aren't that good, even after decades of attempts. A DDG search for 'Boldrin' on that blog shows several articles which reference "Against Intellectual Monopoly", including http://blogs.sciencemag.org/pipeline/archives/2006/04/02/down_with_patents_eh http://blogs.sciencemag.org/pipeline/archives/2006/04/02/dow... . I am too ill-informed to be able to provide any commentary on my own. As I understand it, most of the "me-too" drugs are because multiple drug companies are working with relatively few drug targets. If company X comes out with a drug, and company Y comes out with a 'me-too' drug two years later, most likely what happened was that companies X and Y started looking for a drug some 15-20 years ago against the same target, and found similar compounds. Then when X got its patent, Y had to figure out some way to extract profit from its decade of work, which would otherwise go down the drain. It's not so much that Y "researched" X's existing drug to make a me-too compound. In addition, just because some drug appears to be a "me-too" drug, that doesn't make it useless. People's bodies are different. Quoting from a comment at http://blogs.sciencemag.org/pipeline/archives/2004/08/31/me_too_part_two http://blogs.sciencemag.org/pipeline/archives/2004/08/31/me_... , "Some people may not respond to Zocor but may respond beautifully to Lipitor. Or they may experience less muscle weakness under Crestor." See also http://blogs.sciencemag.org/pipeline/archives/2004/02/29/more_from_the_metoo_front http://blogs.sciencemag.org/pipeline/archives/2004/02/29/mor... .