Alex Tabarrok reports that getting a drug approved by the FDA costs an average $1B. This reduces the flow of new drugs and increases the cost of drugs, especially for rare diseases. Therefore, he suggests that the FDA only require that manufacturers prove that drugs are safe, not that they are efficacious (that is, that they work better than an existing treatments or placebos). He argues, as I have also argued, that under current FDA regulations it will be difficult to develop drugs targeted to specific patient genetic signatures, because such drugs will have extremely small markets.
Tabarrok proposes that the FDA require fewer studies and less data before approving drugs. This would mean that more drugs with little or no benefit would be released. Despite his suggestion that the FDA could still assure us that drugs were safe, his plan would mean that more drugs with significant harms would be released. Even under current FDA guidelines, some serious adverse drug effects are detected only in large trials.
That doesn't mean that Tabarrok is wrong. Finding good drugs is almost entirely a trial and error process. So it is important to try lots of new drugs to discover new ones that work.
So, how would we discover which drugs work? Under Tabarrok's plan we would get less information about a given drug from randomized clinical trials. We would, however, get more information from markets for medical services. There are many medical marketplaces where I expect this would work. These would be markets for drugs, devices, or services that are purchased directly by patients; the classic case being prosthetics (e.g., contact lenses).
There are also many situations where it is difficult to imagine how patient preferences could drive the discovery problem. Consider the problem of chemotherapies tied to genomic signatures, which Tabarrok highlights. A patient offered an option of a chemotherapy tailored to her genome will likely know no similar patients with whom she will share experience. The cancer center may have treated only a handful of similar patients. How will a market detect a signal here?
I would support Tabarrok's plan in a heartbeat if we had an effective system for monitoring and evaluating the effectiveness and safety of medications after their release. This means:
- Routine registration of cases in national databases.
- Routine clinical use of standard outcome measures for diseases, with outcomes reported to the national databases.
- Ongoing comparative effectiveness analyses of de-identified outcomes data, by anyone, to inform clinicians, patients, and insurers about the safety and effectiveness of treatments.
We do not have anything like this. Given our highly decentralized health care system, there are many technical obstacles building a national outcomes monitoring system. It would need to be carefully engineered to protect patient privacy. Finally, it would impose a small cost on every episode of medical care. Sum that over all of us and all our care, and you are talking real money.
Regardless, I think this is the system we need. It would allow us to lower the cost of innovation, as Tabarrok argues. Even more importantly, it would be a foundation for a truly evidence-based medicine.