Thanks to Charlie Hooper for drawing attention to a terrific Wall Street Journal article this week on the FDA's compassionate use program, which allows a small fraction of patients (about 1200) battling deadly diseases to receive access to experimental drugs.
The compassionate use program is designed to allow patients with serious or life threatening diseases to access medicines that are still in development, and haven't been approved for marketing by the FDA. In this case, Dr. Nisha Gupta was infected with Hepatitis C, a very nasty viral infection that, until relatively recently, had few good treatment options. Dr. Gupta became very ill, with the disease eventually causing liver cancer, leading to a liver transplant.
Dr. Gupta petitioned several companies and the FDA for access to experimental therapies, with one company, Bristol Myers Squibb (BMS), eventually granting her and her physician access to daclatasvir. (My fellow blogger Josh Bloom has written quite a bit about the next generation of antiviral therapies for Hep C, which represents a true breakthrough in the field). Today, thanks to BMS, Dr. Gupta is doing much better.
But the WSJ article implies that there is some sharp break between "experimental" medicines and FDA approved drugs.
The reality is far more complex. The public has the false impression that after the FDA approves a product, it is 100% safe and 100% effective. This is simply impossible. The enormous complexity of human biology means that no drug is safe and effective for every patient under every conceivable use - let alone in combination with other drugs the patient may be taking.
And yet pressures from Congress and the public to meet this impossible standard have led the FDA to demand ever more safety and clinical trial data from drug manufacturers, making drug R&D ever more expensive and time consuming. Some commentators have gone so far as to call this Moore's law in reverse, or Eroom's law, where productivity is falling across the industry even as costs for new technologies (like genomic sequencing) plummet.
Source: The Productivity Crisis in Pharmaceutical R&D
The FDA, or at least its senior leadership, has understood this challenge at least since the AIDS epidemic, and has moved to create regulatory "safety valves" that allow some compounds to come to market without the full dossier of safety and efficacy data otherwise required from manufacturers.
Compassionate use is one of those "safety valves" programs. Accelerated approval is another. These programs recognize that patients like Dr. Gupta will die if they don't have access to "experimental" therapies for which there are good - not perfect, but good - reasons to believe that they might help them.
The trouble is that these programs are very small, and (and the case of accelerated approval) have been largely limited to cancer, HIV, and orphan drug indications. It's also no coincidence that these diseases benefit from very well organized, very vocal, and very influential patient's groups. Other diseases - usually slower killers that afflict millions of Americans, aren't so lucky. Think diabetes, obesity, and central nervous systems (CNS) disorders like Alzheimer's.
One study, by the Tufts Center for the Study of Drug Development, found that CNS disorders spend 102 months in clinical review, 40 percent longer than non-CNS drugs.
What do to do about this disparity? And how can we create a better societal approach to balancing the risks and benefits of new medicines that can accelerate innovation?
The evolution of cancer treatment gives us a lens through which to view the future. Many new targeted cancer treatments attack only a small fraction of the true genetic diversity of cancer. Cancer is not one disease, or even one hundred diseases. There may be dozens of different subtypes of even rare cancers like gastrointestinal stromal tumors (GIST), which afflict just a few thousand patients annually.
Take the new, and very effective, drug crizotinib, from Pfizer. Crizotinib is an anaplastic lymphoma kinase (ALK) inhibitor, and the ALK rearrangement is found in about 4% of lung cancers. Now, crizotinib is a great drug for this population, and 4% of a common cancer like lung cancer is still a great many tumors, but what about the other 96%?
We're going to need far more drugs - many, many more drugs - to challenge not just the genetic diversity of cancer but the inevitable development of tumor resistance to targeted therapies. Even relatively "simple" cancers like chronic myelogenous leukemia (CML), with a single driving genetic mutation (BCR-ABL), will eventually develop resistance to powerful targeted drugs.
Gleevec, first approved by the FDA in 2001 may work for CML 5 or 10 years in many patients. But resistance will come. And eventually, patients will develop resistance to follow on versions of Gleevec as well.
So, even with many of the best drugs we have for CML today, in a relatively uncomplicated cancer, what we're really doing is buying time. The trade-off is eminently worth it, because we're talking about years or decades for patients, but we haven't achieved the kind of disease control we have for, say, AIDS.
At the other end of the spectrum, new molecular screening technologies are uncovering prospective targets in cancer much faster than we're coming up with drugs for them.
So the challenge is to put more shots on target, and develop multiple drug cocktails that shut down multiple targets simultaneously and eventually conquer the problem of tumor resistance (again, AIDS is the paradigm here).
Drug companies are very worried about this. Why are they worried? Because the time and cost necessary to bring even targeted therapies to market is still staggering. Even in a banner year - 2011 - the FDA approved only 35 drugs (and not all of them for cancer, obviously). Crizotinib, a miracle drug for some lung cancer patients, took five years to go from lab to patients.
That's blazing speed for any drug, but still far too slow, given the challenge. And think about the need, for a moment.
Not just for Dr. Gupta and a few thousand patients she represents for the compassionate use program, but for the over 500,000 patients who die every year from metastatic cancer; 80,000 from Alzheimer's; and nearly 70,000 from diabetes.
Tweaking the system around the edges - expanding compassionate use, for instance - is not what we need.
What do we need?
Another approach, gaining currency among regulators and drug companies, is the idea of adaptive licensing. Adaptive licensing would allow market access to targeted populations early in the drug testing process, i.e., after basic safety and efficacy testing is completed. Drugs would then be followed in the postmarket environment through electronic medical records.
Adaptive licensing/approval might be followed by ongoing randomized controlled trials to confirm efficacy or uncover rare adverse effects, or outcomes could be validated by using observational methods or targeted diagnostics. In some cases, just comparing the treatment group to the natural history of the disease (for ALS, for instance) might be sufficient for translating adaptive licensing into full approval.
The trade-off, or bargain, is that companies start selling their drugs in very small but targeted patient populations with very high medical need, but then expand the label and indications as data is developed. This would be an iterative digital learning process that breaks down the barriers that currently exist between clinical research and real world patients.
Another way to think of it would be as a "rolling" approach to drug approval where safety and efficacy data would be continually collected in different populations in the "real world" to expand use, restrict use, or withdraw the product.
Obvious hurdles that an adaptive licensing approach would have to overcome would include liability concerns, patent issues (companies will have to be convinced that they FDA won't trap their drug in niche population while the patent life of the drug is eroding), and convincing providers and insurers that adaptive licensing wouldn't foist expensive and unproven medicines on a credulous public.
But I think these are all very tractable problems. The key for success would be shifting from the idea of informed consent, which is now focused patients enrolled in highly selective clinical trials, to an idea of informed choice in a market environment.
Consumers would have to accept more uncertainty in some respects, but it would be in return for greater potential benefits in areas of high unmet medical need.
The FDA's job would also shift from gatekeeper to chief information officer, ensuring that patients and physicians were empowered with the data they need to make smarter choices.
The irony is that we often don't have the information we need to make truly informed choices today. Drugs are tested in highly artificial clinical trials, before they are released into large populations. In fact, the way we test and approve drugs today actually penalizes companies for taking a more stepwise approach to learning about their products, because the patent clock is ticking every minute they are in testing.
In short, the FDA's compassionate use program is a very small tool, for what is a very big problem. We need a new paradigm for thinking about patients and sustaining breakthrough innovations - offering patients less compassion, and better informed choices.